04-UCI with pipelines.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " --- \n",
    "# UCI - Default from Credit Card Clients\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "toc-hr-collapsed": true
   },
   "source": [
    "# Dataset presentation\n",
    "\n",
    "This dataset contains information on default payments, demographic factors, credit data, history of payment, and bill statements of credit card clients in Taiwan from April 2005 to September 2005.\n",
    "\n",
    "It can be found here:\n",
    "https://archive.ics.uci.edu/ml/datasets/default+of+credit+card+clients"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Variables\n",
    "There are 25 variables:\n",
    "\n",
    "* ID: ID of each client\n",
    "* LIMIT_BAL: Amount of given credit in NT dollars (includes individual and family/supplementary credit\n",
    "* SEX: Gender (1=male, 2=female)\n",
    "* EDUCATION: (1=graduate school, 2=university, 3=high school, 4=others, 5=unknown, 6=unknown)\n",
    "* MARRIAGE: Marital status (1=married, 2=single, 3=others)\n",
    "* AGE: Age in years\n",
    "* PAY_0: Repayment status in September, 2005 (-1=pay duly, 1=payment delay for one month, 2=payment delay for two months, ... 8=payment delay for eight months, 9=payment delay for nine months and above)\n",
    "* PAY_2: Repayment status in August, 2005 (scale same as above)\n",
    "* PAY_3: Repayment status in July, 2005 (scale same as above)\n",
    "* PAY_4: Repayment status in June, 2005 (scale same as above)\n",
    "* PAY_5: Repayment status in May, 2005 (scale same as above)\n",
    "* PAY_6: Repayment status in April, 2005 (scale same as above)\n",
    "* BILL_AMT1: Amount of bill statement in September, 2005 (NT dollar)\n",
    "* BILL_AMT2: Amount of bill statement in August, 2005 (NT dollar)\n",
    "* BILL_AMT3: Amount of bill statement in July, 2005 (NT dollar)\n",
    "* BILL_AMT4: Amount of bill statement in June, 2005 (NT dollar)\n",
    "* BILL_AMT5: Amount of bill statement in May, 2005 (NT dollar)\n",
    "* BILL_AMT6: Amount of bill statement in April, 2005 (NT dollar)\n",
    "* PAY_AMT1: Amount of previous payment in September, 2005 (NT dollar)\n",
    "* PAY_AMT2: Amount of previous payment in August, 2005 (NT dollar)\n",
    "* PAY_AMT3: Amount of previous payment in July, 2005 (NT dollar)\n",
    "* PAY_AMT4: Amount of previous payment in June, 2005 (NT dollar)\n",
    "* PAY_AMT5: Amount of previous payment in May, 2005 (NT dollar)\n",
    "* PAY_AMT6: Amount of previous payment in April, 2005 (NT dollar)\n",
    "* default.payment.next.month: Default payment (1=yes, 0=no)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "toc-hr-collapsed": true
   },
   "source": [
    "# Useful imports"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import sklearn\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.set_option('display.max_columns', 120)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set_style(\"darkgrid\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preparation imports"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import GridSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xgboost as xgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 302,
   "metadata": {},
   "outputs": [],
   "source": [
    "import lightgbm as lgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 248,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import catboost"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Metrics imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import recall_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1104,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost\\\\Library\\\\mingw-w64\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost\\\\Library\\\\usr\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost\\\\Library\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost\\\\Scripts',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\envs\\\\boost\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\condabin',\n",
       " 'C:\\\\Program Files\\\\Docker\\\\Docker\\\\Resources\\\\bin',\n",
       " 'C:\\\\ProgramData\\\\Oracle\\\\Java\\\\javapath',\n",
       " 'C:\\\\WINDOWS\\\\system32',\n",
       " 'C:\\\\WINDOWS',\n",
       " 'C:\\\\WINDOWS\\\\System32\\\\Wbem',\n",
       " 'C:\\\\WINDOWS\\\\System32\\\\WindowsPowerShell\\\\v1.0',\n",
       " 'C:\\\\Program Files\\\\PuTTY',\n",
       " 'C:\\\\Program Files\\\\Intel\\\\WiFi\\\\bin',\n",
       " 'C:\\\\Program Files\\\\Common Files\\\\Intel\\\\WirelessCommon',\n",
       " 'C:\\\\Program Files\\\\Git LFS',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\Library\\\\mingw-w64\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\Library\\\\usr\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\Library\\\\bin',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Continuum\\\\anaconda3\\\\Scripts',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Microsoft\\\\WindowsApps',\n",
       " 'C:\\\\Users\\\\twang\\\\AppData\\\\Local\\\\Programs\\\\Git\\\\cmd',\n",
       " 'C:\\\\Program Files\\\\Intel\\\\WiFi\\\\bin',\n",
       " 'C:\\\\Program Files\\\\Common Files\\\\Intel\\\\WirelessCommon']"
      ]
     },
     "execution_count": 1104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "os.environ['PATH'].split(';')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Importing our Data easily!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 250,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = 'UCI_Credit_Card.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.load import load_raw_data\n",
    "from dataprep.load import load_data\n",
    "from dataprep.load import load_data_xy\n",
    "\n",
    "\n",
    "df = load_data(path)\n",
    "X_raw, y_raw = load_data_xy(path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Protocol for pipelined workflow"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "# * Pipeline for lowercasing\n",
    "* pipeline for re-labelling the mislabelled data\n",
    "\n",
    "* pipe for ageBins\n",
    "* pipe for gender x Marriage\n",
    "* pipe for gender x age\n",
    "* pipe for next month's bill\n",
    "---> One output: X_cat, que l'on normalise avec pipe=StandardScaler\n",
    "* pipe for 1-hot\n",
    "* pipe for scaling\n",
    "---> We get our second output X_1hot"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Puis on se doit de définir les dictionnaires / classes que l'on va utiliser pour stocker les modèles, leurs scores selon toutes les métriques"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "models = []\n",
    "models.append(LogisticReg...)\n",
    "\n",
    "scores_accuracy={}\n",
    "scores_recall\n",
    "scores_auc\n",
    "features_importances={}\n",
    "chacun de ces dictionnaires contenant les performances du modèle: {'test': score_test, 'train': score_train}\n",
    "for model in models:\n",
    "    scores_accuracy{[model][train]} = accuracy_score(model.predict(X_test), y_test)\n",
    "    scores_accuracy{[model][test]} = accuracy(model.predict(X_train), y_train)\n",
    "    idem for recall, auc...\n",
    "    features_importances[model] = features_importanes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ainsi, on pourra les capitalizeer facilement"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pipeline bricks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pipeline bricks are pipelines which only do a small elementary change (such as adding a feature, deleting a feature...)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Age binning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>age_bin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>50s</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  pay_1  pay_2  pay_3  pay_4  pay_5  \\\n",
       "0    20000.0       2          2         1      1      1      0      0      0   \n",
       "1   120000.0       2          2         2      0      1      0      0      0   \n",
       "2    90000.0       2          2         2      0      0      0      0      0   \n",
       "3    50000.0       2          2         1      0      0      0      0      0   \n",
       "4    50000.0       1          2         1      0      0      0      0      0   \n",
       "\n",
       "   pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  \\\n",
       "0      0     3913.0     3102.0      689.0        0.0        0.0        0.0   \n",
       "1      1     2682.0     1725.0     2682.0     3272.0     3455.0     3261.0   \n",
       "2      0    29239.0    14027.0    13559.0    14331.0    14948.0    15549.0   \n",
       "3      0    46990.0    48233.0    49291.0    28314.0    28959.0    29547.0   \n",
       "4      0     8617.0     5670.0    35835.0    20940.0    19146.0    19131.0   \n",
       "\n",
       "   pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6 age_bin  \n",
       "0       0.0     689.0       0.0       0.0       0.0       0.0     20s  \n",
       "1       0.0    1000.0    1000.0    1000.0       0.0    2000.0     20s  \n",
       "2    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0     30s  \n",
       "3    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0     30s  \n",
       "4    2000.0   36681.0   10000.0    9000.0     689.0     679.0     50s  "
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.feature_eng import bin_age\n",
    "\n",
    "bin_age(X_raw).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>age_bin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>50s</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  pay_1  pay_2  pay_3  pay_4  pay_5  \\\n",
       "0    20000.0       2          2         1      1      1      0      0      0   \n",
       "1   120000.0       2          2         2      0      1      0      0      0   \n",
       "2    90000.0       2          2         2      0      0      0      0      0   \n",
       "3    50000.0       2          2         1      0      0      0      0      0   \n",
       "4    50000.0       1          2         1      0      0      0      0      0   \n",
       "\n",
       "   pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  \\\n",
       "0      0     3913.0     3102.0      689.0        0.0        0.0        0.0   \n",
       "1      1     2682.0     1725.0     2682.0     3272.0     3455.0     3261.0   \n",
       "2      0    29239.0    14027.0    13559.0    14331.0    14948.0    15549.0   \n",
       "3      0    46990.0    48233.0    49291.0    28314.0    28959.0    29547.0   \n",
       "4      0     8617.0     5670.0    35835.0    20940.0    19146.0    19131.0   \n",
       "\n",
       "   pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6 age_bin  \n",
       "0       0.0     689.0       0.0       0.0       0.0       0.0     20s  \n",
       "1       0.0    1000.0    1000.0    1000.0       0.0    2000.0     20s  \n",
       "2    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0     30s  \n",
       "3    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0     30s  \n",
       "4    2000.0   36681.0   10000.0    9000.0     689.0     679.0     50s  "
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.pipelines import AgeBinAdder\n",
    "\n",
    "age_adder = AgeBinAdder()\n",
    "age_adder.transform(X_raw).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15 ms ± 684 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "age_adder = AgeBinAdder()\n",
    "X = age_adder.transform(X_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15 ms ± 480 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
     ]
    }
   ],
   "source": [
    "%%timeit\n",
    "X = bin_age(X_raw)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ca marche, et ça prend le même temps, au top!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pipeline for gender x Marriage"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>gen_mar</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(2, 1)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>(2, 2)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>(2, 2)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>(2, 1)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>(1, 1)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "  gen_mar  \n",
       "0  (2, 1)  \n",
       "1  (2, 2)  \n",
       "2  (2, 2)  \n",
       "3  (2, 1)  \n",
       "4  (1, 1)  "
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.pipelines import GenderXMarriageAdder\n",
    "\n",
    "gender_x_marriage_adder = GenderXMarriageAdder()\n",
    "gender_x_marriage_adder.transform(X_raw).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw, y_raw = load_data_xy(path)\n",
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>gen_mar</th>\n",
       "      <th>age_bin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>(1, 1)</td>\n",
       "      <td>50s</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  pay_1  pay_2  pay_3  pay_4  pay_5  \\\n",
       "0    20000.0       2          2         1      1      1      0      0      0   \n",
       "1   120000.0       2          2         2      0      1      0      0      0   \n",
       "2    90000.0       2          2         2      0      0      0      0      0   \n",
       "3    50000.0       2          2         1      0      0      0      0      0   \n",
       "4    50000.0       1          2         1      0      0      0      0      0   \n",
       "\n",
       "   pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  \\\n",
       "0      0     3913.0     3102.0      689.0        0.0        0.0        0.0   \n",
       "1      1     2682.0     1725.0     2682.0     3272.0     3455.0     3261.0   \n",
       "2      0    29239.0    14027.0    13559.0    14331.0    14948.0    15549.0   \n",
       "3      0    46990.0    48233.0    49291.0    28314.0    28959.0    29547.0   \n",
       "4      0     8617.0     5670.0    35835.0    20940.0    19146.0    19131.0   \n",
       "\n",
       "   pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6 gen_mar age_bin  \n",
       "0       0.0     689.0       0.0       0.0       0.0       0.0  (2, 1)     20s  \n",
       "1       0.0    1000.0    1000.0    1000.0       0.0    2000.0  (2, 2)     20s  \n",
       "2    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  (2, 2)     30s  \n",
       "3    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  (2, 1)     30s  \n",
       "4    2000.0   36681.0   10000.0    9000.0     689.0     679.0  (1, 1)     50s  "
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "age_adder.transform(gender_x_marriage_adder.transform(X_raw)).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pipeline for gender x agebin"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.pipelines import GenderXAgeBinAdder"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pipeline for predicted balance amount"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 854,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_raw, y_raw = load_data_xy(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 855,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 855,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 856,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>pred_bill_amt0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3909.683795</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2678.148995</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>29177.529785</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>46995.038110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>8605.584386</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "   pred_bill_amt0  \n",
       "0     3909.683795  \n",
       "1     2678.148995  \n",
       "2    29177.529785  \n",
       "3    46995.038110  \n",
       "4     8605.584386  "
      ]
     },
     "execution_count": 856,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.pipelines import NextBillAdder\n",
    "\n",
    "next_bill_adder = NextBillAdder()\n",
    "next_bill_adder.transform(X_raw).head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 857,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 857,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Toutes mes pipelines marchent bien, et ne modifient pas l'input mais renvoient une deep copy modifiée!\n",
    "\n",
    "Reste à définir les pipelines encapsulantes, qui comprennent en plus:\n",
    "* 1-hot encoder si nécessaire\n",
    "* standard scaler"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Les pipelines finales"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ces pipelines \"finales\" prennent en entrée le path, ou bien le DF, et sortent des données normalisées, et peuvent être 1-hot encoded ou non"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1-hot encoded Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 861,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "\n",
    "pipeline_1hot = Pipeline([\n",
    "    (\"age_bin\", AgeBinAdder()),\n",
    "    (\"gender_x_marriage\", GenderXMarriageAdder()),\n",
    "    (\"gender_x_age_bin\", GenderXAgeBinAdder()),\n",
    "    (\"next_bill\", NextBillAdder()),\n",
    "    (\"1-hot\", OneHotEncoder(categories='auto'))\n",
    "     \n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 862,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 862,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.compose import ColumnTransformer\n",
    "\n",
    "ct = ColumnTransformer(\n",
    "    [('oh_enc', OneHotEncoder(sparse=False), [1, 2, 3]),],  # the column numbers I want to apply this to\n",
    "    remainder='passthrough'  # This leaves the rest of my columns in place\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30000, 23)"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "limit_bal    float64\n",
       "gender         int64\n",
       "education      int64\n",
       "marriage       int64\n",
       "age            int64\n",
       "pay_1          int64\n",
       "pay_2          int64\n",
       "pay_3          int64\n",
       "pay_4          int64\n",
       "pay_5          int64\n",
       "pay_6          int64\n",
       "bill_amt1    float64\n",
       "bill_amt2    float64\n",
       "bill_amt3    float64\n",
       "bill_amt4    float64\n",
       "bill_amt5    float64\n",
       "bill_amt6    float64\n",
       "pay_amt1     float64\n",
       "pay_amt2     float64\n",
       "pay_amt3     float64\n",
       "pay_amt4     float64\n",
       "pay_amt5     float64\n",
       "pay_amt6     float64\n",
       "dtype: object"
      ]
     },
     "execution_count": 114,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\preprocessing\\_encoders.py:415: FutureWarning: The handling of integer data will change in version 0.22. Currently, the categories are determined based on the range [0, max(values)], while in the future they will be determined based on the unique values.\n",
      "If you want the future behaviour and silence this warning, you can specify \"categories='auto'\".\n",
      "In case you used a LabelEncoder before this OneHotEncoder to convert the categories to integers, then you can now use the OneHotEncoder directly.\n",
      "  warnings.warn(msg, FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(30000, 29)"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ct.fit_transform(X_raw).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "CategoricalWarrior(attribute_names=['gender', 'education', 'marriage'])"
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.pipelines import CategoricalWarrior\n",
    "cat_war = CategoricalWarrior(['gender', 'education', 'marriage'])\n",
    "cat_war.fit(X_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5, 29)"
      ]
     },
     "execution_count": 127,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_war.transform(X_raw).head().shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>gender_2</th>\n",
       "      <th>gender_1</th>\n",
       "      <th>education_2</th>\n",
       "      <th>education_1</th>\n",
       "      <th>education_3</th>\n",
       "      <th>education_4</th>\n",
       "      <th>marriage_1</th>\n",
       "      <th>marriage_2</th>\n",
       "      <th>marriage_3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  age  pay_1  pay_2  pay_3  pay_4  pay_5  pay_6  bill_amt1  \\\n",
       "0    20000.0   24      1      1      0      0      0      0     3913.0   \n",
       "1   120000.0   26      0      1      0      0      0      1     2682.0   \n",
       "2    90000.0   34      0      0      0      0      0      0    29239.0   \n",
       "3    50000.0   37      0      0      0      0      0      0    46990.0   \n",
       "4    50000.0   57      0      0      0      0      0      0     8617.0   \n",
       "\n",
       "   bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  pay_amt1  pay_amt2  \\\n",
       "0     3102.0      689.0        0.0        0.0        0.0       0.0     689.0   \n",
       "1     1725.0     2682.0     3272.0     3455.0     3261.0       0.0    1000.0   \n",
       "2    14027.0    13559.0    14331.0    14948.0    15549.0    1518.0    1500.0   \n",
       "3    48233.0    49291.0    28314.0    28959.0    29547.0    2000.0    2019.0   \n",
       "4     5670.0    35835.0    20940.0    19146.0    19131.0    2000.0   36681.0   \n",
       "\n",
       "   pay_amt3  pay_amt4  pay_amt5  pay_amt6  gender_2  gender_1  education_2  \\\n",
       "0       0.0       0.0       0.0       0.0         1         0            1   \n",
       "1    1000.0    1000.0       0.0    2000.0         1         0            1   \n",
       "2    1000.0    1000.0    1000.0    5000.0         1         0            1   \n",
       "3    1200.0    1100.0    1069.0    1000.0         1         0            1   \n",
       "4   10000.0    9000.0     689.0     679.0         0         1            1   \n",
       "\n",
       "   education_1  education_3  education_4  marriage_1  marriage_2  marriage_3  \n",
       "0            0            0            0           1           0           0  \n",
       "1            0            0            0           0           1           0  \n",
       "2            0            0            0           0           1           0  \n",
       "3            0            0            0           1           0           0  \n",
       "4            0            0            0           1           0           0  "
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_war.transform(X_raw).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'gender': [2, 1], 'education': [2, 1, 3, 4], 'marriage': [1, 2, 3]}"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_war.categoricals"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['limit_bal', 'age', 'pay_1', 'pay_2', 'pay_3', 'pay_4', 'pay_5',\n",
       "       'pay_6', 'bill_amt1', 'bill_amt2', 'bill_amt3', 'bill_amt4',\n",
       "       'bill_amt5', 'bill_amt6', 'pay_amt1', 'pay_amt2', 'pay_amt3',\n",
       "       'pay_amt4', 'pay_amt5', 'pay_amt6', 'gender_2', 'gender_1',\n",
       "       'education_2', 'education_1', 'education_3', 'education_4',\n",
       "       'marriage_1', 'marriage_2', 'marriage_3'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_war.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2    14030\n",
       "1    10585\n",
       "3     4917\n",
       "4      468\n",
       "Name: education, dtype: int64"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw['education'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 863,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "\n",
    "pipeline_1hot = Pipeline([\n",
    "    (\"age_bin\", AgeBinAdder()),\n",
    "    (\"gender_x_marriage\", GenderXMarriageAdder()),\n",
    "    (\"gender_x_age_bin\", GenderXAgeBinAdder()),\n",
    "    (\"next_bill\", NextBillAdder()),\n",
    "    (\"cat_war\",  CategoricalWarrior(['gender', 'education', 'marriage']))\n",
    "     \n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 864,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>pred_bill_amt0</th>\n",
       "      <th>gender_2</th>\n",
       "      <th>gender_1</th>\n",
       "      <th>education_2</th>\n",
       "      <th>education_1</th>\n",
       "      <th>education_3</th>\n",
       "      <th>education_4</th>\n",
       "      <th>marriage_1</th>\n",
       "      <th>marriage_2</th>\n",
       "      <th>marriage_3</th>\n",
       "      <th>age_bin_20s</th>\n",
       "      <th>age_bin_30s</th>\n",
       "      <th>age_bin_40s</th>\n",
       "      <th>age_bin_50s</th>\n",
       "      <th>age_bin_60+</th>\n",
       "      <th>gen_mar_(1, 1)</th>\n",
       "      <th>gen_mar_(1, 2)</th>\n",
       "      <th>gen_mar_(1, 3)</th>\n",
       "      <th>gen_mar_(2, 1)</th>\n",
       "      <th>gen_mar_(2, 2)</th>\n",
       "      <th>gen_mar_(2, 3)</th>\n",
       "      <th>gen_ageBin_(1, '20s')</th>\n",
       "      <th>gen_ageBin_(1, '30s')</th>\n",
       "      <th>gen_ageBin_(1, '40s')</th>\n",
       "      <th>gen_ageBin_(1, '50s')</th>\n",
       "      <th>gen_ageBin_(1, '60+')</th>\n",
       "      <th>gen_ageBin_(2, '20s')</th>\n",
       "      <th>gen_ageBin_(2, '30s')</th>\n",
       "      <th>gen_ageBin_(2, '40s')</th>\n",
       "      <th>gen_ageBin_(2, '50s')</th>\n",
       "      <th>gen_ageBin_(2, '60+')</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3909.683795</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2678.148995</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>29177.529785</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>46995.038110</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>8605.584386</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  age  pay_1  pay_2  pay_3  pay_4  pay_5  pay_6  bill_amt1  \\\n",
       "0    20000.0   24      1      1      0      0      0      0     3913.0   \n",
       "1   120000.0   26      0      1      0      0      0      1     2682.0   \n",
       "2    90000.0   34      0      0      0      0      0      0    29239.0   \n",
       "3    50000.0   37      0      0      0      0      0      0    46990.0   \n",
       "4    50000.0   57      0      0      0      0      0      0     8617.0   \n",
       "\n",
       "   bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  pay_amt1  pay_amt2  \\\n",
       "0     3102.0      689.0        0.0        0.0        0.0       0.0     689.0   \n",
       "1     1725.0     2682.0     3272.0     3455.0     3261.0       0.0    1000.0   \n",
       "2    14027.0    13559.0    14331.0    14948.0    15549.0    1518.0    1500.0   \n",
       "3    48233.0    49291.0    28314.0    28959.0    29547.0    2000.0    2019.0   \n",
       "4     5670.0    35835.0    20940.0    19146.0    19131.0    2000.0   36681.0   \n",
       "\n",
       "   pay_amt3  pay_amt4  pay_amt5  pay_amt6  pred_bill_amt0  gender_2  gender_1  \\\n",
       "0       0.0       0.0       0.0       0.0     3909.683795         1         0   \n",
       "1    1000.0    1000.0       0.0    2000.0     2678.148995         1         0   \n",
       "2    1000.0    1000.0    1000.0    5000.0    29177.529785         1         0   \n",
       "3    1200.0    1100.0    1069.0    1000.0    46995.038110         1         0   \n",
       "4   10000.0    9000.0     689.0     679.0     8605.584386         0         1   \n",
       "\n",
       "   education_2  education_1  education_3  education_4  marriage_1  marriage_2  \\\n",
       "0            1            0            0            0           1           0   \n",
       "1            1            0            0            0           0           1   \n",
       "2            1            0            0            0           0           1   \n",
       "3            1            0            0            0           1           0   \n",
       "4            1            0            0            0           1           0   \n",
       "\n",
       "   marriage_3  age_bin_20s  age_bin_30s  age_bin_40s  age_bin_50s  \\\n",
       "0           0            1            0            0            0   \n",
       "1           0            1            0            0            0   \n",
       "2           0            0            1            0            0   \n",
       "3           0            0            1            0            0   \n",
       "4           0            0            0            0            1   \n",
       "\n",
       "   age_bin_60+  gen_mar_(1, 1)  gen_mar_(1, 2)  gen_mar_(1, 3)  \\\n",
       "0            0               0               0               0   \n",
       "1            0               0               0               0   \n",
       "2            0               0               0               0   \n",
       "3            0               0               0               0   \n",
       "4            0               1               0               0   \n",
       "\n",
       "   gen_mar_(2, 1)  gen_mar_(2, 2)  gen_mar_(2, 3)  gen_ageBin_(1, '20s')  \\\n",
       "0               1               0               0                      0   \n",
       "1               0               1               0                      0   \n",
       "2               0               1               0                      0   \n",
       "3               1               0               0                      0   \n",
       "4               0               0               0                      0   \n",
       "\n",
       "   gen_ageBin_(1, '30s')  gen_ageBin_(1, '40s')  gen_ageBin_(1, '50s')  \\\n",
       "0                      0                      0                      0   \n",
       "1                      0                      0                      0   \n",
       "2                      0                      0                      0   \n",
       "3                      0                      0                      0   \n",
       "4                      0                      0                      1   \n",
       "\n",
       "   gen_ageBin_(1, '60+')  gen_ageBin_(2, '20s')  gen_ageBin_(2, '30s')  \\\n",
       "0                      0                      1                      0   \n",
       "1                      0                      1                      0   \n",
       "2                      0                      0                      1   \n",
       "3                      0                      0                      1   \n",
       "4                      0                      0                      0   \n",
       "\n",
       "   gen_ageBin_(2, '40s')  gen_ageBin_(2, '50s')  gen_ageBin_(2, '60+')  \n",
       "0                      0                      0                      0  \n",
       "1                      0                      0                      0  \n",
       "2                      0                      0                      0  \n",
       "3                      0                      0                      0  \n",
       "4                      0                      0                      0  "
      ]
     },
     "execution_count": 864,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pipeline_1hot.fit_transform(X_raw).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "On a enfin un transformer qui marche!"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Puis, il faut splitter en test/train/val\n",
    "\n",
    "Et enfin, scaler chacun de nos sets en utilisant les valeurs moyennes et sigma2 de training set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  education  marriage  age  pay_1  pay_2  pay_3  pay_4  pay_5  \\\n",
       "0    20000.0          2         1   24      1      1      0      0      0   \n",
       "1   120000.0          2         2   26      0      1      0      0      0   \n",
       "2    90000.0          2         2   34      0      0      0      0      0   \n",
       "3    50000.0          2         1   37      0      0      0      0      0   \n",
       "4    50000.0          2         1   57      0      0      0      0      0   \n",
       "\n",
       "   pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  \\\n",
       "0      0     3913.0     3102.0      689.0        0.0        0.0        0.0   \n",
       "1      1     2682.0     1725.0     2682.0     3272.0     3455.0     3261.0   \n",
       "2      0    29239.0    14027.0    13559.0    14331.0    14948.0    15549.0   \n",
       "3      0    46990.0    48233.0    49291.0    28314.0    28959.0    29547.0   \n",
       "4      0     8617.0     5670.0    35835.0    20940.0    19146.0    19131.0   \n",
       "\n",
       "   pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from dataprep.pipelines import ColumnDropper\n",
    "\n",
    "ColumnDropper('gender').transform(X_raw).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Testing our full-Log Reg Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "metadata": {},
   "outputs": [],
   "source": [
    "log_clf = LogisticRegression(C=0.1, \n",
    "                             solver='liblinear',\n",
    "                             penalty='l2',\n",
    "                             class_weight='balanced', \n",
    "                             random_state=42, \n",
    "                             n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [],
   "source": [
    "prep_pipe = Pipeline([\n",
    "    ('prep_1hot', pipeline_1hot),\n",
    "    ('scaler', StandardScaler()),\n",
    "    #('lr', log_clf)\n",
    "    #('rf', RandomForestClassifier())\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-1.13672015, -1.24601985,  1.84394071, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ],\n",
       "       [-0.3659805 , -1.02904717, -0.54231679, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ],\n",
       "       [-0.59720239, -0.16115646, -0.54231679, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ],\n",
       "       ...,\n",
       "       [-1.05964618,  0.16430256,  1.84394071, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ],\n",
       "       [-0.67427636,  0.59824792,  1.84394071, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ],\n",
       "       [-0.90549825,  1.14067961, -0.54231679, ..., -0.35858507,\n",
       "        -0.18694037, -0.0732252 ]])"
      ]
     },
     "execution_count": 226,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prep_pipe.fit_transform(X_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {},
   "outputs": [],
   "source": [
    "log_pipe = Pipeline([\n",
    "    ('prep_1hot', pipeline_1hot),\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('lr', log_clf)\n",
    "    #('rf', RandomForestClassifier())\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:1544: UserWarning: 'n_jobs' > 1 does not have any effect when 'solver' is set to 'liblinear'. Got 'n_jobs' = 8.\n",
      "  \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "         steps=[('prep_1hot',\n",
       "                 Pipeline(memory=None,\n",
       "                          steps=[('age_bin', AgeBinAdder()),\n",
       "                                 ('gender_x_marriage', GenderXMarriageAdder()),\n",
       "                                 ('gender_x_age_bin', GenderXAgeBinAdder()),\n",
       "                                 ('cat_war',\n",
       "                                  CategoricalWarrior(attribute_names=['gender',\n",
       "                                                                      'education',\n",
       "                                                                      'marriage']))],\n",
       "                          verbose=False)),\n",
       "                ('scaler',\n",
       "                 StandardScaler(copy=True, with_mean=True, with_std=True)),\n",
       "                ('lr',\n",
       "                 LogisticRegression(C=0.1, class_weight='balanced', dual=False,\n",
       "                                    fit_intercept=True, intercept_scaling=1,\n",
       "                                    l1_ratio=None, max_iter=100,\n",
       "                                    multi_class='warn', n_jobs=-1, penalty='l2',\n",
       "                                    random_state=42, solver='liblinear',\n",
       "                                    tol=0.0001, verbose=0, warm_start=False))],\n",
       "         verbose=False)"
      ]
     },
     "execution_count": 228,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_pipe.fit(X_raw, np.ravel(y_raw))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.83      0.85     23364\n",
      "           1       0.48      0.58      0.53      6636\n",
      "\n",
      "    accuracy                           0.77     30000\n",
      "   macro avg       0.68      0.70      0.69     30000\n",
      "weighted avg       0.79      0.77      0.78     30000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_raw, log_pipe.predict(X_raw)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Bulk-defining XGBClassifier-hot encoded models\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 865,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.ensemble import AdaBoostClassifier\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 955,
   "metadata": {},
   "outputs": [],
   "source": [
    "models = []\n",
    "\n",
    "dico_models = {}\n",
    "dico_pipelined_models = {}\n",
    "\n",
    "dico_results = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 956,
   "metadata": {},
   "outputs": [],
   "source": [
    "dico_models['log'] = log_clf = LogisticRegression(C=0.1, \n",
    "                             solver='liblinear',\n",
    "                             penalty='l2',\n",
    "                             class_weight='balanced', \n",
    "                             random_state=42, \n",
    "                             n_jobs=-1)\n",
    "dico_models['svm'] =  SVC(gamma='auto', C=1, class_weight='balanced')\n",
    "dico_models['tree'] = DecisionTreeClassifier(criterion='entropy', \n",
    "                                  random_state=42, \n",
    "                                 max_leaf_nodes=5,\n",
    "                                 class_weight='balanced')\n",
    "dico_models['forest'] = RandomForestClassifier(n_estimators=500, \n",
    "                                 max_leaf_nodes=10, \n",
    "                                 n_jobs=-1, \n",
    "                                 class_weight='balanced',\n",
    "                                random_state=42)\n",
    "\n",
    "dico_models['adaboost'] = AdaBoostClassifier()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 957,
   "metadata": {},
   "outputs": [],
   "source": [
    "for (name, model) in dico_models.items():\n",
    "    dico_pipelined_models[name] = Pipeline([\n",
    "        ('prep_1hot', pipeline_1hot),\n",
    "        ('scaler', StandardScaler()),\n",
    "        (name, model)\n",
    "        #('rf', RandomForestClassifier())\n",
    "    ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 958,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "log\n",
      "svm\n",
      "tree\n",
      "forest\n",
      "adaboost\n"
     ]
    }
   ],
   "source": [
    "for i in dico_pipelined_models.keys():\n",
    "    print (i)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 959,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_raw, y_raw = load_data_xy(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 960,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:1544: UserWarning: 'n_jobs' > 1 does not have any effect when 'solver' is set to 'liblinear'. Got 'n_jobs' = 8.\n",
      "  \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Log\n",
      "--------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.83      0.85     23364\n",
      "           1       0.48      0.58      0.53      6636\n",
      "\n",
      "    accuracy                           0.77     30000\n",
      "   macro avg       0.68      0.70      0.69     30000\n",
      "weighted avg       0.79      0.77      0.78     30000\n",
      "\n",
      "\n",
      "Svm\n",
      "--------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.81      0.85     23364\n",
      "           1       0.49      0.63      0.55      6636\n",
      "\n",
      "    accuracy                           0.77     30000\n",
      "   macro avg       0.69      0.72      0.70     30000\n",
      "weighted avg       0.80      0.77      0.78     30000\n",
      "\n",
      "\n",
      "Tree\n",
      "--------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.83      0.85     23364\n",
      "           1       0.48      0.55      0.51      6636\n",
      "\n",
      "    accuracy                           0.77     30000\n",
      "   macro avg       0.67      0.69      0.68     30000\n",
      "weighted avg       0.78      0.77      0.77     30000\n",
      "\n",
      "\n",
      "Forest\n",
      "--------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.79      0.83     23364\n",
      "           1       0.46      0.63      0.53      6636\n",
      "\n",
      "    accuracy                           0.75     30000\n",
      "   macro avg       0.67      0.71      0.68     30000\n",
      "weighted avg       0.79      0.75      0.77     30000\n",
      "\n",
      "\n",
      "Adaboost\n",
      "--------\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.83      0.95      0.89     23364\n",
      "           1       0.64      0.32      0.42      6636\n",
      "\n",
      "    accuracy                           0.81     30000\n",
      "   macro avg       0.73      0.63      0.65     30000\n",
      "weighted avg       0.79      0.81      0.78     30000\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for (name, pipe) in dico_pipelined_models.items():\n",
    "    pipe.fit(X_raw, np.ravel(y_raw))\n",
    "    print (name.capitalize())\n",
    "    print ('--------')\n",
    "    dico_results[name] = classification_report(y_raw, pipe.predict(X_raw))\n",
    "    print (dico_results[name])\n",
    "    print ('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1152,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\utils\\validation.py:724: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().\n",
      "  y = column_or_1d(y, warn=True)\n",
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:1544: UserWarning: 'n_jobs' > 1 does not have any effect when 'solver' is set to 'liblinear'. Got 'n_jobs' = 8.\n",
      "  \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=0.1, class_weight='balanced', dual=False,\n",
       "                   fit_intercept=True, intercept_scaling=1, l1_ratio=None,\n",
       "                   max_iter=100, multi_class='warn', n_jobs=-1, penalty='l2',\n",
       "                   random_state=42, solver='liblinear', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 1152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_clf.fit(X_raw, y_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1153,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.51      0.64     23364\n",
      "           1       0.29      0.72      0.42      6636\n",
      "\n",
      "    accuracy                           0.56     30000\n",
      "   macro avg       0.58      0.61      0.53     30000\n",
      "weighted avg       0.74      0.56      0.59     30000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_raw, log_clf.predict(X_raw)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SMOTE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1138,
   "metadata": {},
   "outputs": [],
   "source": [
    "from imblearn.over_sampling import SMOTE\n",
    "\n",
    "smt = SMOTE(random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1139,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>30000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.221200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.415062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            default\n",
       "count  30000.000000\n",
       "mean       0.221200\n",
       "std        0.415062\n",
       "min        0.000000\n",
       "25%        0.000000\n",
       "50%        0.000000\n",
       "75%        0.000000\n",
       "max        1.000000"
      ]
     },
     "execution_count": 1139,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_raw.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1140,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_smote, y_smote = smt.fit_sample(X_raw, np.ravel(y_raw))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1141,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([23364, 23364], dtype=int64)"
      ]
     },
     "execution_count": 1141,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.bincount(y_smote)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1142,
   "metadata": {},
   "outputs": [],
   "source": [
    "models_SMOTE = []\n",
    "\n",
    "dico_models_SMOTE = {}\n",
    "dico_pipelined_models_SMOTE = {}\n",
    "\n",
    "dico_results_SMOTE = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1143,
   "metadata": {},
   "outputs": [],
   "source": [
    "dico_models_SMOTE['log'] = log_clf = LogisticRegression(C=0.1, \n",
    "                             solver='liblinear',\n",
    "                             penalty='l2',\n",
    "                             class_weight='balanced', \n",
    "                             random_state=42, \n",
    "                             n_jobs=-1)\n",
    "dico_models_SMOTE['svm'] =  SVC(gamma='auto', C=1, class_weight='balanced')\n",
    "dico_models_SMOTE['tree'] = DecisionTreeClassifier(criterion='entropy', \n",
    "                                  random_state=42, \n",
    "                                 max_leaf_nodes=5,\n",
    "                                 class_weight='balanced')\n",
    "dico_models_SMOTE['forest'] = RandomForestClassifier(n_estimators=500, \n",
    "                                 max_leaf_nodes=10, \n",
    "                                 n_jobs=-1, \n",
    "                                 class_weight='balanced',\n",
    "                                random_state=42)\n",
    "\n",
    "dico_models_SMOTE['adaboost'] = AdaBoostClassifier()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1144,
   "metadata": {},
   "outputs": [],
   "source": [
    "for (name, model) in dico_models_SMOTE.items():\n",
    "    dico_pipelined_models_SMOTE[name] = Pipeline([\n",
    "        ('prep_1hot', pipeline_1hot),\n",
    "        ('scaler', StandardScaler()),\n",
    "        (name, model)\n",
    "        #('rf', RandomForestClassifier())\n",
    "    ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1145,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "log\n",
      "svm\n",
      "tree\n",
      "forest\n",
      "adaboost\n"
     ]
    }
   ],
   "source": [
    "for i in dico_pipelined_models_SMOTE.keys():\n",
    "    print (i)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1146,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_raw, y_raw = load_data_xy(path)\n",
    "X_SMOTE, y_SMOTE = smt.fit_sample(X_raw, np.ravel(y_raw))\n",
    "X_SMOTE = pd.DataFrame(X_SMOTE, columns=X_raw.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1147,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:1544: UserWarning: 'n_jobs' > 1 does not have any effect when 'solver' is set to 'liblinear'. Got 'n_jobs' = 8.\n",
      "  \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=0.1, class_weight='balanced', dual=False,\n",
       "                   fit_intercept=True, intercept_scaling=1, l1_ratio=None,\n",
       "                   max_iter=100, multi_class='warn', n_jobs=-1, penalty='l2',\n",
       "                   random_state=42, solver='liblinear', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 1147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_clf.fit(X_SMOTE, np.ravel(y_SMOTE))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.66      0.52      0.58     23364\n",
      "           1       0.61      0.73      0.66     23364\n",
      "\n",
      "    accuracy                           0.63     46728\n",
      "   macro avg       0.63      0.63      0.62     46728\n",
      "weighted avg       0.63      0.63      0.62     46728\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_SMOTE, log_clf.predict(X_SMOTE)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.52      0.65     23364\n",
      "           1       0.30      0.71      0.42      6636\n",
      "\n",
      "    accuracy                           0.57     30000\n",
      "   macro avg       0.58      0.62      0.54     30000\n",
      "weighted avg       0.74      0.57      0.60     30000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_raw, log_clf.predict(X_raw)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1151,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>30000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.221200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.415062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            default\n",
       "count  30000.000000\n",
       "mean       0.221200\n",
       "std        0.415062\n",
       "min        0.000000\n",
       "25%        0.000000\n",
       "50%        0.000000\n",
       "75%        0.000000\n",
       "max        1.000000"
      ]
     },
     "execution_count": 1151,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_raw.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "toc-hr-collapsed": false
   },
   "source": [
    "# Splitting Train/test/val"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Stratified"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "toc-hr-collapsed": false
   },
   "source": [
    "# Feature Engineering"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Age Binning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 874,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.feature_eng import bin_age\n",
    "\n",
    "df = bin_age(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 875,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>default</th>\n",
       "      <th>age_bin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50s</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "   default age_bin  \n",
       "0        1     20s  \n",
       "1        1     20s  \n",
       "2        0     30s  \n",
       "3        0     30s  \n",
       "4        0     50s  "
      ]
     },
     "execution_count": 875,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## New category Gender x Marriage"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 876,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.feature_eng import gender_x_marriage\n",
    "\n",
    "df = gender_x_marriage(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 877,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>default</th>\n",
       "      <th>age_bin</th>\n",
       "      <th>gen_mar</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 1)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 2)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 2)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 1)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50s</td>\n",
       "      <td>(1, 1)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "   default age_bin gen_mar  \n",
       "0        1     20s  (2, 1)  \n",
       "1        1     20s  (2, 2)  \n",
       "2        0     30s  (2, 2)  \n",
       "3        0     30s  (2, 1)  \n",
       "4        0     50s  (1, 1)  "
      ]
     },
     "execution_count": 877,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## New category Gender x Age"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 878,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.feature_eng import gender_x_agebin\n",
    "\n",
    "df = gender_x_agebin(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 879,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>default</th>\n",
       "      <th>age_bin</th>\n",
       "      <th>gen_mar</th>\n",
       "      <th>gen_ageBin</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>(2, 20s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>(2, 20s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>(2, 30s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>(2, 30s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50s</td>\n",
       "      <td>(1, 1)</td>\n",
       "      <td>(1, 50s)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "   default age_bin gen_mar gen_ageBin  \n",
       "0        1     20s  (2, 1)   (2, 20s)  \n",
       "1        1     20s  (2, 2)   (2, 20s)  \n",
       "2        0     30s  (2, 2)   (2, 30s)  \n",
       "3        0     30s  (2, 1)   (2, 30s)  \n",
       "4        0     50s  (1, 1)   (1, 50s)  "
      ]
     },
     "execution_count": 879,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's add a linear Bill_amt0 prediction to our features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 880,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.feature_eng import next_bill_amt\n",
    "    \n",
    "df = next_bill_amt(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 881,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>default</th>\n",
       "      <th>age_bin</th>\n",
       "      <th>gen_mar</th>\n",
       "      <th>gen_ageBin</th>\n",
       "      <th>pred_bill_amt0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>(2, 20s)</td>\n",
       "      <td>3909.683795</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>20s</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>(2, 20s)</td>\n",
       "      <td>2678.148995</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 2)</td>\n",
       "      <td>(2, 30s)</td>\n",
       "      <td>29177.529785</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>30s</td>\n",
       "      <td>(2, 1)</td>\n",
       "      <td>(2, 30s)</td>\n",
       "      <td>46995.038110</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>50s</td>\n",
       "      <td>(1, 1)</td>\n",
       "      <td>(1, 50s)</td>\n",
       "      <td>8605.584386</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \\\n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0   \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0   \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0   \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0   \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0   \n",
       "\n",
       "   default age_bin gen_mar gen_ageBin  pred_bill_amt0  \n",
       "0        1     20s  (2, 1)   (2, 20s)     3909.683795  \n",
       "1        1     20s  (2, 2)   (2, 20s)     2678.148995  \n",
       "2        0     30s  (2, 2)   (2, 30s)    29177.529785  \n",
       "3        0     30s  (2, 1)   (2, 30s)    46995.038110  \n",
       "4        0     50s  (1, 1)   (1, 50s)     8605.584386  "
      ]
     },
     "execution_count": 881,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preparing our data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## One-hot encoding our categorical data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 882,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cat = df.copy(deep=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 883,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.get_dummies(df, columns=['gender', 'education', 'marriage', 'gen_mar', 'gen_ageBin', 'age_bin'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 884,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "      <th>default</th>\n",
       "      <th>pred_bill_amt0</th>\n",
       "      <th>gender_1</th>\n",
       "      <th>gender_2</th>\n",
       "      <th>education_1</th>\n",
       "      <th>education_2</th>\n",
       "      <th>education_3</th>\n",
       "      <th>education_4</th>\n",
       "      <th>marriage_1</th>\n",
       "      <th>marriage_2</th>\n",
       "      <th>marriage_3</th>\n",
       "      <th>gen_mar_(1, 1)</th>\n",
       "      <th>gen_mar_(1, 2)</th>\n",
       "      <th>gen_mar_(1, 3)</th>\n",
       "      <th>gen_mar_(2, 1)</th>\n",
       "      <th>gen_mar_(2, 2)</th>\n",
       "      <th>gen_mar_(2, 3)</th>\n",
       "      <th>gen_ageBin_(1, '20s')</th>\n",
       "      <th>gen_ageBin_(1, '30s')</th>\n",
       "      <th>gen_ageBin_(1, '40s')</th>\n",
       "      <th>gen_ageBin_(1, '50s')</th>\n",
       "      <th>gen_ageBin_(1, '60+')</th>\n",
       "      <th>gen_ageBin_(2, '20s')</th>\n",
       "      <th>gen_ageBin_(2, '30s')</th>\n",
       "      <th>gen_ageBin_(2, '40s')</th>\n",
       "      <th>gen_ageBin_(2, '50s')</th>\n",
       "      <th>gen_ageBin_(2, '60+')</th>\n",
       "      <th>age_bin_20s</th>\n",
       "      <th>age_bin_30s</th>\n",
       "      <th>age_bin_40s</th>\n",
       "      <th>age_bin_50s</th>\n",
       "      <th>age_bin_60+</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>3909.683795</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2678.148995</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>29177.529785</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0</td>\n",
       "      <td>46995.038110</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>0</td>\n",
       "      <td>8605.584386</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  age  pay_1  pay_2  pay_3  pay_4  pay_5  pay_6  bill_amt1  \\\n",
       "0    20000.0   24      1      1      0      0      0      0     3913.0   \n",
       "1   120000.0   26      0      1      0      0      0      1     2682.0   \n",
       "2    90000.0   34      0      0      0      0      0      0    29239.0   \n",
       "3    50000.0   37      0      0      0      0      0      0    46990.0   \n",
       "4    50000.0   57      0      0      0      0      0      0     8617.0   \n",
       "\n",
       "   bill_amt2  bill_amt3  bill_amt4  bill_amt5  bill_amt6  pay_amt1  pay_amt2  \\\n",
       "0     3102.0      689.0        0.0        0.0        0.0       0.0     689.0   \n",
       "1     1725.0     2682.0     3272.0     3455.0     3261.0       0.0    1000.0   \n",
       "2    14027.0    13559.0    14331.0    14948.0    15549.0    1518.0    1500.0   \n",
       "3    48233.0    49291.0    28314.0    28959.0    29547.0    2000.0    2019.0   \n",
       "4     5670.0    35835.0    20940.0    19146.0    19131.0    2000.0   36681.0   \n",
       "\n",
       "   pay_amt3  pay_amt4  pay_amt5  pay_amt6  default  pred_bill_amt0  gender_1  \\\n",
       "0       0.0       0.0       0.0       0.0        1     3909.683795         0   \n",
       "1    1000.0    1000.0       0.0    2000.0        1     2678.148995         0   \n",
       "2    1000.0    1000.0    1000.0    5000.0        0    29177.529785         0   \n",
       "3    1200.0    1100.0    1069.0    1000.0        0    46995.038110         0   \n",
       "4   10000.0    9000.0     689.0     679.0        0     8605.584386         1   \n",
       "\n",
       "   gender_2  education_1  education_2  education_3  education_4  marriage_1  \\\n",
       "0         1            0            1            0            0           1   \n",
       "1         1            0            1            0            0           0   \n",
       "2         1            0            1            0            0           0   \n",
       "3         1            0            1            0            0           1   \n",
       "4         0            0            1            0            0           1   \n",
       "\n",
       "   marriage_2  marriage_3  gen_mar_(1, 1)  gen_mar_(1, 2)  gen_mar_(1, 3)  \\\n",
       "0           0           0               0               0               0   \n",
       "1           1           0               0               0               0   \n",
       "2           1           0               0               0               0   \n",
       "3           0           0               0               0               0   \n",
       "4           0           0               1               0               0   \n",
       "\n",
       "   gen_mar_(2, 1)  gen_mar_(2, 2)  gen_mar_(2, 3)  gen_ageBin_(1, '20s')  \\\n",
       "0               1               0               0                      0   \n",
       "1               0               1               0                      0   \n",
       "2               0               1               0                      0   \n",
       "3               1               0               0                      0   \n",
       "4               0               0               0                      0   \n",
       "\n",
       "   gen_ageBin_(1, '30s')  gen_ageBin_(1, '40s')  gen_ageBin_(1, '50s')  \\\n",
       "0                      0                      0                      0   \n",
       "1                      0                      0                      0   \n",
       "2                      0                      0                      0   \n",
       "3                      0                      0                      0   \n",
       "4                      0                      0                      1   \n",
       "\n",
       "   gen_ageBin_(1, '60+')  gen_ageBin_(2, '20s')  gen_ageBin_(2, '30s')  \\\n",
       "0                      0                      1                      0   \n",
       "1                      0                      1                      0   \n",
       "2                      0                      0                      1   \n",
       "3                      0                      0                      1   \n",
       "4                      0                      0                      0   \n",
       "\n",
       "   gen_ageBin_(2, '40s')  gen_ageBin_(2, '50s')  gen_ageBin_(2, '60+')  \\\n",
       "0                      0                      0                      0   \n",
       "1                      0                      0                      0   \n",
       "2                      0                      0                      0   \n",
       "3                      0                      0                      0   \n",
       "4                      0                      0                      0   \n",
       "\n",
       "   age_bin_20s  age_bin_30s  age_bin_40s  age_bin_50s  age_bin_60+  \n",
       "0            1            0            0            0            0  \n",
       "1            1            0            0            0            0  \n",
       "2            0            1            0            0            0  \n",
       "3            0            1            0            0            0  \n",
       "4            0            0            0            1            0  "
      ]
     },
     "execution_count": 884,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Splitting test/train/val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 885,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "df_train, df_val = train_test_split(df, test_size=6000, stratify=df.default, random_state=42)\n",
    "df_train, df_test = train_test_split(df_train, test_size=6000, stratify=df_train.default, random_state=42)\n",
    "\n",
    "df_train_cat, df_val_cat = train_test_split(df_cat, test_size=6000, stratify=df_cat.default, random_state=42)\n",
    "df_train_cat, df_test_cat = train_test_split(df_train_cat, test_size=6000, stratify=df_train_cat.default, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Splitting into (X,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 886,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataprep.load import df2xy\n",
    "\n",
    "X_train, y_train = df2xy(df_train, 'default')\n",
    "X_test, y_test = df2xy(df_test, 'default')\n",
    "X_val, y_val = df2xy(df_val, 'default')\n",
    "\n",
    "X_train_cat, y_train_cat = df2xy(df_train_cat, 'default')\n",
    "X_test_cat, y_test_cat= df2xy(df_test_cat, 'default')\n",
    "X_val_cat, y_va_cat = df2xy(df_val_cat, 'default')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 887,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "      <th>default</th>\n",
       "      <th>default</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>18000.000000</td>\n",
       "      <td>6000.000000</td>\n",
       "      <td>6000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.221222</td>\n",
       "      <td>0.221167</td>\n",
       "      <td>0.221167</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.415081</td>\n",
       "      <td>0.415067</td>\n",
       "      <td>0.415067</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            default      default      default\n",
       "count  18000.000000  6000.000000  6000.000000\n",
       "mean       0.221222     0.221167     0.221167\n",
       "std        0.415081     0.415067     0.415067\n",
       "min        0.000000     0.000000     0.000000\n",
       "25%        0.000000     0.000000     0.000000\n",
       "50%        0.000000     0.000000     0.000000\n",
       "75%        0.000000     0.000000     0.000000\n",
       "max        1.000000     1.000000     1.000000"
      ]
     },
     "execution_count": 887,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.concat([y_train,y_test, y_val], axis=1).describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It looks like our data is well stratified: the proportions of defaults in our 3 data sets are all ~22%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 888,
   "metadata": {},
   "outputs": [],
   "source": [
    "colonnes_dummies = list(X_train.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Normalizing our training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 889,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "scaler = StandardScaler()\n",
    "X_train_norm = scaler.fit_transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 890,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test_norm = scaler.transform(X_test)\n",
    "X_val_norm = scaler.transform(X_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training our models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 891,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\linear_model\\logistic.py:1544: UserWarning: 'n_jobs' > 1 does not have any effect when 'solver' is set to 'liblinear'. Got 'n_jobs' = 8.\n",
      "  \" = {}.\".format(effective_n_jobs(self.n_jobs)))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=0.1, class_weight='balanced', dual=False,\n",
       "                   fit_intercept=True, intercept_scaling=1, l1_ratio=None,\n",
       "                   max_iter=100, multi_class='warn', n_jobs=-1, penalty='l2',\n",
       "                   random_state=42, solver='liblinear', tol=0.0001, verbose=0,\n",
       "                   warm_start=False)"
      ]
     },
     "execution_count": 891,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "log_reg = LogisticRegression(C=0.1, \n",
    "                             solver='liblinear',\n",
    "                             penalty='l2',\n",
    "                             class_weight='balanced', \n",
    "                             random_state=42, \n",
    "                             n_jobs=-1)\n",
    "log_reg.fit(X_train_norm, np.ravel(y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 892,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.7722222222222223\n",
      "recall:  0.5878955298844801\n",
      "F1:  0.5331359599180141\n",
      "--TEST--\n",
      "accuracy:  0.772\n",
      "recall:  0.5719668425018839\n",
      "F1:  0.525987525987526\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = log_reg.predict(X_train_norm)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = log_reg.predict(X_test_norm)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 893,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.88      0.82      0.85     14018\n",
      "     default       0.49      0.59      0.53      3982\n",
      "\n",
      "    accuracy                           0.77     18000\n",
      "   macro avg       0.68      0.71      0.69     18000\n",
      "weighted avg       0.79      0.77      0.78     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 894,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.87      0.83      0.85      4673\n",
      "     default       0.49      0.57      0.53      1327\n",
      "\n",
      "    accuracy                           0.77      6000\n",
      "   macro avg       0.68      0.70      0.69      6000\n",
      "weighted avg       0.79      0.77      0.78      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 895,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7062391046483323\n",
      "0.7003853044095125\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 896,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.23152069 -0.0180804   0.56492595  0.06567299  0.14241522  0.10603199\n",
      "   0.0794715   0.11424631 -0.02561625  0.03825804  0.11141664  0.07288891\n",
      "   0.01616337 -0.11118334 -0.1324497  -0.16970286  0.00446142 -0.02496143\n",
      "  -0.02695221 -0.03540794 -0.0253706   0.01129104 -0.01129104  0.00370817\n",
      "   0.02958255 -0.00778328 -0.10944303  0.02025779 -0.02041717  0.00091522\n",
      "   0.02417586 -0.01461419  0.03565215  0.00209567 -0.0089222  -0.02656937\n",
      "  -0.00857949  0.0186665   0.0143472  -0.00719204 -0.01288672 -0.00364241\n",
      "  -0.02068536 -0.00438551  0.02398944  0.02191974 -0.00915574 -0.00392061\n",
      "   0.00665432  0.0128138   0.00494339]]\n"
     ]
    }
   ],
   "source": [
    "print(log_reg.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 897,
   "metadata": {},
   "outputs": [],
   "source": [
    "coefs=np.array(log_reg.coef_[0])\n",
    "abs_coefs = np.abs(coefs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 898,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_importances_log = pd.DataFrame(list(zip(colonnes_dummies, coefs, abs_coefs)),\n",
    "                                  columns=['Feature', 'Coef', 'Abs Coef'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 899,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Feature</th>\n",
       "      <th>Coef</th>\n",
       "      <th>Abs Coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>pay_1</td>\n",
       "      <td>0.564926</td>\n",
       "      <td>0.564926</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>limit_bal</td>\n",
       "      <td>-0.231521</td>\n",
       "      <td>0.231521</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>pay_amt2</td>\n",
       "      <td>-0.169703</td>\n",
       "      <td>0.169703</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>pay_3</td>\n",
       "      <td>0.142415</td>\n",
       "      <td>0.142415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>pay_amt1</td>\n",
       "      <td>-0.132450</td>\n",
       "      <td>0.132450</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>pay_6</td>\n",
       "      <td>0.114246</td>\n",
       "      <td>0.114246</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>bill_amt3</td>\n",
       "      <td>0.111417</td>\n",
       "      <td>0.111417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>bill_amt6</td>\n",
       "      <td>-0.111183</td>\n",
       "      <td>0.111183</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>education_4</td>\n",
       "      <td>-0.109443</td>\n",
       "      <td>0.109443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>pay_4</td>\n",
       "      <td>0.106032</td>\n",
       "      <td>0.106032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>pay_5</td>\n",
       "      <td>0.079471</td>\n",
       "      <td>0.079471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>bill_amt4</td>\n",
       "      <td>0.072889</td>\n",
       "      <td>0.072889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>pay_2</td>\n",
       "      <td>0.065673</td>\n",
       "      <td>0.065673</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>bill_amt2</td>\n",
       "      <td>0.038258</td>\n",
       "      <td>0.038258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>gen_mar_(1, 3)</td>\n",
       "      <td>0.035652</td>\n",
       "      <td>0.035652</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>pay_amt6</td>\n",
       "      <td>-0.035408</td>\n",
       "      <td>0.035408</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>education_2</td>\n",
       "      <td>0.029583</td>\n",
       "      <td>0.029583</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>pay_amt5</td>\n",
       "      <td>-0.026952</td>\n",
       "      <td>0.026952</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>gen_mar_(2, 3)</td>\n",
       "      <td>-0.026569</td>\n",
       "      <td>0.026569</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>bill_amt1</td>\n",
       "      <td>-0.025616</td>\n",
       "      <td>0.025616</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>pred_bill_amt0</td>\n",
       "      <td>-0.025371</td>\n",
       "      <td>0.025371</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>pay_amt4</td>\n",
       "      <td>-0.024961</td>\n",
       "      <td>0.024961</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>gen_mar_(1, 1)</td>\n",
       "      <td>0.024176</td>\n",
       "      <td>0.024176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>gen_ageBin_(2, '50s')</td>\n",
       "      <td>0.023989</td>\n",
       "      <td>0.023989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>gen_ageBin_(2, '60+')</td>\n",
       "      <td>0.021920</td>\n",
       "      <td>0.021920</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>gen_ageBin_(2, '30s')</td>\n",
       "      <td>-0.020685</td>\n",
       "      <td>0.020685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>marriage_2</td>\n",
       "      <td>-0.020417</td>\n",
       "      <td>0.020417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>marriage_1</td>\n",
       "      <td>0.020258</td>\n",
       "      <td>0.020258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>gen_ageBin_(1, '30s')</td>\n",
       "      <td>0.018666</td>\n",
       "      <td>0.018666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>age</td>\n",
       "      <td>-0.018080</td>\n",
       "      <td>0.018080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>bill_amt5</td>\n",
       "      <td>0.016163</td>\n",
       "      <td>0.016163</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>gen_mar_(1, 2)</td>\n",
       "      <td>-0.014614</td>\n",
       "      <td>0.014614</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>gen_ageBin_(1, '40s')</td>\n",
       "      <td>0.014347</td>\n",
       "      <td>0.014347</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>gen_ageBin_(1, '60+')</td>\n",
       "      <td>-0.012887</td>\n",
       "      <td>0.012887</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>age_bin_50s</td>\n",
       "      <td>0.012814</td>\n",
       "      <td>0.012814</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>gender_2</td>\n",
       "      <td>-0.011291</td>\n",
       "      <td>0.011291</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>gender_1</td>\n",
       "      <td>0.011291</td>\n",
       "      <td>0.011291</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>age_bin_20s</td>\n",
       "      <td>-0.009156</td>\n",
       "      <td>0.009156</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>gen_mar_(2, 2)</td>\n",
       "      <td>-0.008922</td>\n",
       "      <td>0.008922</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>gen_ageBin_(1, '20s')</td>\n",
       "      <td>-0.008579</td>\n",
       "      <td>0.008579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>education_3</td>\n",
       "      <td>-0.007783</td>\n",
       "      <td>0.007783</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>gen_ageBin_(1, '50s')</td>\n",
       "      <td>-0.007192</td>\n",
       "      <td>0.007192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>age_bin_40s</td>\n",
       "      <td>0.006654</td>\n",
       "      <td>0.006654</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>age_bin_60+</td>\n",
       "      <td>0.004943</td>\n",
       "      <td>0.004943</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>pay_amt3</td>\n",
       "      <td>0.004461</td>\n",
       "      <td>0.004461</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>gen_ageBin_(2, '40s')</td>\n",
       "      <td>-0.004386</td>\n",
       "      <td>0.004386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>age_bin_30s</td>\n",
       "      <td>-0.003921</td>\n",
       "      <td>0.003921</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>education_1</td>\n",
       "      <td>0.003708</td>\n",
       "      <td>0.003708</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>gen_ageBin_(2, '20s')</td>\n",
       "      <td>-0.003642</td>\n",
       "      <td>0.003642</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>gen_mar_(2, 1)</td>\n",
       "      <td>0.002096</td>\n",
       "      <td>0.002096</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>marriage_3</td>\n",
       "      <td>0.000915</td>\n",
       "      <td>0.000915</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  Feature      Coef  Abs Coef\n",
       "2                   pay_1  0.564926  0.564926\n",
       "0               limit_bal -0.231521  0.231521\n",
       "15               pay_amt2 -0.169703  0.169703\n",
       "4                   pay_3  0.142415  0.142415\n",
       "14               pay_amt1 -0.132450  0.132450\n",
       "7                   pay_6  0.114246  0.114246\n",
       "10              bill_amt3  0.111417  0.111417\n",
       "13              bill_amt6 -0.111183  0.111183\n",
       "26            education_4 -0.109443  0.109443\n",
       "5                   pay_4  0.106032  0.106032\n",
       "6                   pay_5  0.079471  0.079471\n",
       "11              bill_amt4  0.072889  0.072889\n",
       "3                   pay_2  0.065673  0.065673\n",
       "9               bill_amt2  0.038258  0.038258\n",
       "32         gen_mar_(1, 3)  0.035652  0.035652\n",
       "19               pay_amt6 -0.035408  0.035408\n",
       "24            education_2  0.029583  0.029583\n",
       "18               pay_amt5 -0.026952  0.026952\n",
       "35         gen_mar_(2, 3) -0.026569  0.026569\n",
       "8               bill_amt1 -0.025616  0.025616\n",
       "20         pred_bill_amt0 -0.025371  0.025371\n",
       "17               pay_amt4 -0.024961  0.024961\n",
       "30         gen_mar_(1, 1)  0.024176  0.024176\n",
       "44  gen_ageBin_(2, '50s')  0.023989  0.023989\n",
       "45  gen_ageBin_(2, '60+')  0.021920  0.021920\n",
       "42  gen_ageBin_(2, '30s') -0.020685  0.020685\n",
       "28             marriage_2 -0.020417  0.020417\n",
       "27             marriage_1  0.020258  0.020258\n",
       "37  gen_ageBin_(1, '30s')  0.018666  0.018666\n",
       "1                     age -0.018080  0.018080\n",
       "12              bill_amt5  0.016163  0.016163\n",
       "31         gen_mar_(1, 2) -0.014614  0.014614\n",
       "38  gen_ageBin_(1, '40s')  0.014347  0.014347\n",
       "40  gen_ageBin_(1, '60+') -0.012887  0.012887\n",
       "49            age_bin_50s  0.012814  0.012814\n",
       "22               gender_2 -0.011291  0.011291\n",
       "21               gender_1  0.011291  0.011291\n",
       "46            age_bin_20s -0.009156  0.009156\n",
       "34         gen_mar_(2, 2) -0.008922  0.008922\n",
       "36  gen_ageBin_(1, '20s') -0.008579  0.008579\n",
       "25            education_3 -0.007783  0.007783\n",
       "39  gen_ageBin_(1, '50s') -0.007192  0.007192\n",
       "48            age_bin_40s  0.006654  0.006654\n",
       "50            age_bin_60+  0.004943  0.004943\n",
       "16               pay_amt3  0.004461  0.004461\n",
       "43  gen_ageBin_(2, '40s') -0.004386  0.004386\n",
       "47            age_bin_30s -0.003921  0.003921\n",
       "23            education_1  0.003708  0.003708\n",
       "41  gen_ageBin_(2, '20s') -0.003642  0.003642\n",
       "33         gen_mar_(2, 1)  0.002096  0.002096\n",
       "29             marriage_3  0.000915  0.000915"
      ]
     },
     "execution_count": 899,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_importances_log.sort_values(by=['Abs Coef'], ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hence, it appears that our most important variables seem to be:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 900,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['pay_1',\n",
       " 'limit_bal',\n",
       " 'pay_amt2',\n",
       " 'pay_3',\n",
       " 'pay_amt1',\n",
       " 'pay_6',\n",
       " 'bill_amt3',\n",
       " 'bill_amt6',\n",
       " 'education_4']"
      ]
     },
     "execution_count": 900,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(feature_importances_log.sort_values(by=['Abs Coef'], ascending=False)['Feature'])[:9]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can note that they are the same as before we did our little feature engineering...\n",
    "😅"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 901,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=1, cache_size=200, class_weight='balanced', coef0=0.0,\n",
       "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',\n",
       "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "    tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 901,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.svm import SVC\n",
    "\n",
    "svc_clf = SVC(gamma='auto', C=1, class_weight='balanced')\n",
    "svc_clf.fit(X_train_norm, np.ravel(y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 902,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.7728888888888888\n",
      "recall:  0.6464088397790055\n",
      "F1:  0.5573841489822434\n",
      "--TEST--\n",
      "accuracy:  0.7596666666666667\n",
      "recall:  0.5825169555388093\n",
      "F1:  0.5174029451137886\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = svc_clf.predict(X_train_norm)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = svc_clf.predict(X_test_norm)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 903,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.89      0.81      0.85     14018\n",
      "     default       0.49      0.65      0.56      3982\n",
      "\n",
      "    accuracy                           0.77     18000\n",
      "   macro avg       0.69      0.73      0.70     18000\n",
      "weighted avg       0.80      0.77      0.78     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 904,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.87      0.81      0.84      4673\n",
      "     default       0.47      0.58      0.52      1327\n",
      "\n",
      "    accuracy                           0.76      6000\n",
      "   macro avg       0.67      0.70      0.68      6000\n",
      "weighted avg       0.78      0.76      0.77      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 905,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.727613037381299\n",
      "0.6962445680754181\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 906,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeClassifier(class_weight='balanced', criterion='entropy',\n",
       "                       max_depth=None, max_features=None, max_leaf_nodes=5,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=1, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, presort=False,\n",
       "                       random_state=42, splitter='best')"
      ]
     },
     "execution_count": 906,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "tree_clf = DecisionTreeClassifier(criterion='entropy', \n",
    "                                  random_state=42, \n",
    "                                 max_leaf_nodes=5,\n",
    "                                 class_weight='balanced')\n",
    "tree_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 907,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.7675555555555555\n",
      "recall:  0.5595178302360623\n",
      "F1:  0.5157407407407408\n",
      "--TEST--\n",
      "accuracy:  0.7683333333333333\n",
      "recall:  0.5395629238884703\n",
      "F1:  0.5074415308291991\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = tree_clf.predict(X_train)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = tree_clf.predict(X_test)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 908,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.87      0.83      0.85     14018\n",
      "     default       0.48      0.56      0.52      3982\n",
      "\n",
      "    accuracy                           0.77     18000\n",
      "   macro avg       0.67      0.69      0.68     18000\n",
      "weighted avg       0.78      0.77      0.77     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 909,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.86      0.83      0.85      4673\n",
      "     default       0.48      0.54      0.51      1327\n",
      "\n",
      "    accuracy                           0.77      6000\n",
      "   macro avg       0.67      0.69      0.68      6000\n",
      "weighted avg       0.78      0.77      0.77      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's visualize our Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 910,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.tree import export_graphviz\n",
    "\n",
    "export_graphviz(tree_clf, \n",
    "                out_file='tree_credit_default_featureEngineered.dot',\n",
    "                feature_names=list(X_train.columns),\n",
    "                class_names=['no-default', 'default'],\n",
    "                rounded=True,\n",
    "                filled=True\n",
    "               )"
   ]
  },
  {
   "attachments": {
    "DecisionTreeViz-CreditDefault-FeatureEngineered.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![DecisionTreeViz-CreditDefault-FeatureEngineered.png](attachment:DecisionTreeViz-CreditDefault-FeatureEngineered.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 911,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6930846391870852\n",
      "0.6864302956698932\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 912,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_importance_tree = pd.DataFrame(list(zip(X_train.columns, tree_clf.feature_importances_)), \n",
    "                   columns=['Feature', 'Importance']).sort_values(by=['Importance'], ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hence, it appears that our most important variables seem to be:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 913,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['pay_1',\n",
       " 'pred_bill_amt0',\n",
       " 'pay_amt2',\n",
       " 'pay_4',\n",
       " 'bill_amt5',\n",
       " 'bill_amt6',\n",
       " 'pay_2',\n",
       " 'pay_3',\n",
       " 'pay_5']"
      ]
     },
     "execution_count": 913,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(feature_importance_tree.sort_values(by=['Importance'], ascending=False)['Feature'])[:9]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 914,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight='balanced',\n",
       "                       criterion='gini', max_depth=None, max_features='auto',\n",
       "                       max_leaf_nodes=10, min_impurity_decrease=0.0,\n",
       "                       min_impurity_split=None, min_samples_leaf=1,\n",
       "                       min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "                       n_estimators=500, n_jobs=-1, oob_score=False,\n",
       "                       random_state=42, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 914,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "rnd_clf = RandomForestClassifier(n_estimators=500, \n",
    "                                 max_leaf_nodes=10, \n",
    "                                 n_jobs=-1, \n",
    "                                 class_weight='balanced',\n",
    "                                random_state=42)\n",
    "rnd_clf.fit(X_train, np.ravel(y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 915,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.7576111111111111\n",
      "recall:  0.6396283274736313\n",
      "F1:  0.5386486200697896\n",
      "--TEST--\n",
      "accuracy:  0.7535\n",
      "recall:  0.6186887716654107\n",
      "F1:  0.5261134251842358\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = rnd_clf.predict(X_train)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = rnd_clf.predict(X_test)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 916,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.89      0.79      0.84     14018\n",
      "     default       0.47      0.64      0.54      3982\n",
      "\n",
      "    accuracy                           0.76     18000\n",
      "   macro avg       0.68      0.72      0.69     18000\n",
      "weighted avg       0.79      0.76      0.77     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 917,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.88      0.79      0.83      4673\n",
      "     default       0.46      0.62      0.53      1327\n",
      "\n",
      "    accuracy                           0.75      6000\n",
      "   macro avg       0.67      0.71      0.68      6000\n",
      "weighted avg       0.79      0.75      0.77      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 918,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7153770115039723\n",
      "0.7052356762243168\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 919,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4.46489286e-02, 1.23128374e-03, 2.91193230e-01, 1.59302556e-01,\n",
       "       1.10024610e-01, 7.24129164e-02, 6.74886600e-02, 5.74248919e-02,\n",
       "       1.29684312e-02, 1.00570904e-02, 7.03886654e-03, 6.95496555e-03,\n",
       "       4.83369300e-03, 3.87491992e-03, 3.84083219e-02, 2.92041986e-02,\n",
       "       2.66284104e-02, 1.48127262e-02, 1.09934875e-02, 1.33938655e-02,\n",
       "       1.39080772e-02, 7.15181610e-05, 1.53449989e-04, 1.40404925e-03,\n",
       "       2.63913097e-04, 1.01446270e-04, 5.33689700e-04, 7.65587427e-05,\n",
       "       1.67638229e-05, 0.00000000e+00, 5.34485330e-05, 0.00000000e+00,\n",
       "       5.60053613e-05, 1.00028015e-05, 7.51238041e-05, 0.00000000e+00,\n",
       "       0.00000000e+00, 2.50308099e-05, 0.00000000e+00, 0.00000000e+00,\n",
       "       0.00000000e+00, 0.00000000e+00, 2.23856451e-04, 1.48471882e-05,\n",
       "       2.92299240e-05, 0.00000000e+00, 0.00000000e+00, 5.65470634e-05,\n",
       "       9.76045277e-06, 2.06278722e-05, 0.00000000e+00])"
      ]
     },
     "execution_count": 919,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rnd_clf.feature_importances_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 920,
   "metadata": {},
   "outputs": [],
   "source": [
    "coefs=np.array(rnd_clf.feature_importances_)\n",
    "abs_coefs = np.abs(coefs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 921,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_importances_rf = pd.DataFrame(list(zip(colonnes_dummies, coefs, abs_coefs)),\n",
    "                                  columns=['Feature', 'Coef', 'Abs Coef'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 922,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Feature</th>\n",
       "      <th>Coef</th>\n",
       "      <th>Abs Coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>pay_1</td>\n",
       "      <td>0.291193</td>\n",
       "      <td>0.291193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>pay_2</td>\n",
       "      <td>0.159303</td>\n",
       "      <td>0.159303</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>pay_3</td>\n",
       "      <td>0.110025</td>\n",
       "      <td>0.110025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>pay_4</td>\n",
       "      <td>0.072413</td>\n",
       "      <td>0.072413</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>pay_5</td>\n",
       "      <td>0.067489</td>\n",
       "      <td>0.067489</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>pay_6</td>\n",
       "      <td>0.057425</td>\n",
       "      <td>0.057425</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>limit_bal</td>\n",
       "      <td>0.044649</td>\n",
       "      <td>0.044649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>pay_amt1</td>\n",
       "      <td>0.038408</td>\n",
       "      <td>0.038408</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>pay_amt2</td>\n",
       "      <td>0.029204</td>\n",
       "      <td>0.029204</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>pay_amt3</td>\n",
       "      <td>0.026628</td>\n",
       "      <td>0.026628</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>pay_amt4</td>\n",
       "      <td>0.014813</td>\n",
       "      <td>0.014813</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>pred_bill_amt0</td>\n",
       "      <td>0.013908</td>\n",
       "      <td>0.013908</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>pay_amt6</td>\n",
       "      <td>0.013394</td>\n",
       "      <td>0.013394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>bill_amt1</td>\n",
       "      <td>0.012968</td>\n",
       "      <td>0.012968</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>pay_amt5</td>\n",
       "      <td>0.010993</td>\n",
       "      <td>0.010993</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>bill_amt2</td>\n",
       "      <td>0.010057</td>\n",
       "      <td>0.010057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>bill_amt3</td>\n",
       "      <td>0.007039</td>\n",
       "      <td>0.007039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>bill_amt4</td>\n",
       "      <td>0.006955</td>\n",
       "      <td>0.006955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>bill_amt5</td>\n",
       "      <td>0.004834</td>\n",
       "      <td>0.004834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>bill_amt6</td>\n",
       "      <td>0.003875</td>\n",
       "      <td>0.003875</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>education_1</td>\n",
       "      <td>0.001404</td>\n",
       "      <td>0.001404</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>age</td>\n",
       "      <td>0.001231</td>\n",
       "      <td>0.001231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>education_4</td>\n",
       "      <td>0.000534</td>\n",
       "      <td>0.000534</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>education_2</td>\n",
       "      <td>0.000264</td>\n",
       "      <td>0.000264</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>gen_ageBin_(2, '30s')</td>\n",
       "      <td>0.000224</td>\n",
       "      <td>0.000224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>gender_2</td>\n",
       "      <td>0.000153</td>\n",
       "      <td>0.000153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>education_3</td>\n",
       "      <td>0.000101</td>\n",
       "      <td>0.000101</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>marriage_1</td>\n",
       "      <td>0.000077</td>\n",
       "      <td>0.000077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>gen_mar_(2, 2)</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>0.000075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>gender_1</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>0.000072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>age_bin_30s</td>\n",
       "      <td>0.000057</td>\n",
       "      <td>0.000057</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>gen_mar_(1, 3)</td>\n",
       "      <td>0.000056</td>\n",
       "      <td>0.000056</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>gen_mar_(1, 1)</td>\n",
       "      <td>0.000053</td>\n",
       "      <td>0.000053</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>gen_ageBin_(2, '50s')</td>\n",
       "      <td>0.000029</td>\n",
       "      <td>0.000029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>gen_ageBin_(1, '30s')</td>\n",
       "      <td>0.000025</td>\n",
       "      <td>0.000025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>age_bin_50s</td>\n",
       "      <td>0.000021</td>\n",
       "      <td>0.000021</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>marriage_2</td>\n",
       "      <td>0.000017</td>\n",
       "      <td>0.000017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>gen_ageBin_(2, '40s')</td>\n",
       "      <td>0.000015</td>\n",
       "      <td>0.000015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>gen_mar_(2, 1)</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>age_bin_40s</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>0.000010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>age_bin_20s</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>gen_ageBin_(2, '60+')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>gen_ageBin_(1, '50s')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>gen_ageBin_(2, '20s')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>gen_ageBin_(1, '60+')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>gen_ageBin_(1, '40s')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>gen_ageBin_(1, '20s')</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>gen_mar_(2, 3)</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>gen_mar_(1, 2)</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>marriage_3</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50</th>\n",
       "      <td>age_bin_60+</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  Feature      Coef  Abs Coef\n",
       "2                   pay_1  0.291193  0.291193\n",
       "3                   pay_2  0.159303  0.159303\n",
       "4                   pay_3  0.110025  0.110025\n",
       "5                   pay_4  0.072413  0.072413\n",
       "6                   pay_5  0.067489  0.067489\n",
       "7                   pay_6  0.057425  0.057425\n",
       "0               limit_bal  0.044649  0.044649\n",
       "14               pay_amt1  0.038408  0.038408\n",
       "15               pay_amt2  0.029204  0.029204\n",
       "16               pay_amt3  0.026628  0.026628\n",
       "17               pay_amt4  0.014813  0.014813\n",
       "20         pred_bill_amt0  0.013908  0.013908\n",
       "19               pay_amt6  0.013394  0.013394\n",
       "8               bill_amt1  0.012968  0.012968\n",
       "18               pay_amt5  0.010993  0.010993\n",
       "9               bill_amt2  0.010057  0.010057\n",
       "10              bill_amt3  0.007039  0.007039\n",
       "11              bill_amt4  0.006955  0.006955\n",
       "12              bill_amt5  0.004834  0.004834\n",
       "13              bill_amt6  0.003875  0.003875\n",
       "23            education_1  0.001404  0.001404\n",
       "1                     age  0.001231  0.001231\n",
       "26            education_4  0.000534  0.000534\n",
       "24            education_2  0.000264  0.000264\n",
       "42  gen_ageBin_(2, '30s')  0.000224  0.000224\n",
       "22               gender_2  0.000153  0.000153\n",
       "25            education_3  0.000101  0.000101\n",
       "27             marriage_1  0.000077  0.000077\n",
       "34         gen_mar_(2, 2)  0.000075  0.000075\n",
       "21               gender_1  0.000072  0.000072\n",
       "47            age_bin_30s  0.000057  0.000057\n",
       "32         gen_mar_(1, 3)  0.000056  0.000056\n",
       "30         gen_mar_(1, 1)  0.000053  0.000053\n",
       "44  gen_ageBin_(2, '50s')  0.000029  0.000029\n",
       "37  gen_ageBin_(1, '30s')  0.000025  0.000025\n",
       "49            age_bin_50s  0.000021  0.000021\n",
       "28             marriage_2  0.000017  0.000017\n",
       "43  gen_ageBin_(2, '40s')  0.000015  0.000015\n",
       "33         gen_mar_(2, 1)  0.000010  0.000010\n",
       "48            age_bin_40s  0.000010  0.000010\n",
       "46            age_bin_20s  0.000000  0.000000\n",
       "45  gen_ageBin_(2, '60+')  0.000000  0.000000\n",
       "39  gen_ageBin_(1, '50s')  0.000000  0.000000\n",
       "41  gen_ageBin_(2, '20s')  0.000000  0.000000\n",
       "40  gen_ageBin_(1, '60+')  0.000000  0.000000\n",
       "38  gen_ageBin_(1, '40s')  0.000000  0.000000\n",
       "36  gen_ageBin_(1, '20s')  0.000000  0.000000\n",
       "35         gen_mar_(2, 3)  0.000000  0.000000\n",
       "31         gen_mar_(1, 2)  0.000000  0.000000\n",
       "29             marriage_3  0.000000  0.000000\n",
       "50            age_bin_60+  0.000000  0.000000"
      ]
     },
     "execution_count": 922,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_importances_rf.sort_values(by=['Abs Coef'], ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# To plot later"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hence, it appears that our most important variables seem to be:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 923,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['pay_1',\n",
       " 'pay_2',\n",
       " 'pay_3',\n",
       " 'pay_4',\n",
       " 'pay_5',\n",
       " 'pay_6',\n",
       " 'limit_bal',\n",
       " 'pay_amt1',\n",
       " 'pay_amt2']"
      ]
     },
     "execution_count": 923,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(feature_importances_rf.sort_values(by=['Abs Coef'], ascending=False)['Feature'])[:9]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our new predicted bill_amt0 seems to be more important than bill_amt1, and the previous ones, which is a good sign!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Boostings models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### AdaBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 944,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "AdaBoostClassifier(algorithm='SAMME.R', base_estimator=None, learning_rate=1.0,\n",
       "                   n_estimators=10, random_state=None)"
      ]
     },
     "execution_count": 944,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "\n",
    "ada_clf = AdaBoostClassifier(n_estimators=10)\n",
    "ada_clf.fit(X_train_norm, np.ravel(y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 945,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TEST--\n",
      "accuracy:  0.8037222222222222\n",
      "recall:  0.35585133098945254\n",
      "F1:  0.4451075859902623\n",
      "--TEST--\n",
      "accuracy:  0.803\n",
      "recall:  0.3519216277317257\n",
      "F1:  0.441398865784499\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = ada_clf.predict(X_train_norm)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = ada_clf.predict(X_test_norm)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 946,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.84      0.93      0.88     14018\n",
      "     default       0.59      0.36      0.45      3982\n",
      "\n",
      "    accuracy                           0.80     18000\n",
      "   macro avg       0.71      0.64      0.66     18000\n",
      "weighted avg       0.78      0.80      0.78     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 947,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.83      0.93      0.88      4673\n",
      "     default       0.59      0.35      0.44      1327\n",
      "\n",
      "    accuracy                           0.80      6000\n",
      "   macro avg       0.71      0.64      0.66      6000\n",
      "weighted avg       0.78      0.80      0.78      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 948,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6433986288275841\n",
      "0.6415075718371874\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ces résultats overfittent clairemnt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1069,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,\n",
       "              colsample_bynode=1, colsample_bytree=1, gamma=0,\n",
       "              learning_rate=0.1, max_delta_step=0, max_depth=3,\n",
       "              min_child_weight=1, missing=None, n_estimators=100, n_jobs=1,\n",
       "              nthread=None, objective='binary:logistic', random_state=0,\n",
       "              reg_alpha=0, reg_lambda=1, scale_pos_weight=5, seed=None,\n",
       "              silent=None, subsample=1, verbosity=1)"
      ]
     },
     "execution_count": 1069,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb_clf = xgb.XGBClassifier(scale_pos_weight=5)\n",
    "xgb_clf.fit(X_train, np.ravel(y_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1070,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TEST--\n",
      "accuracy:  0.7060555555555555\n",
      "recall:  0.7694625816172778\n",
      "F1:  0.536649443909274\n",
      "--TEST--\n",
      "accuracy:  0.691\n",
      "recall:  0.7226827430293896\n",
      "F1:  0.5084835630965004\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = xgb_clf.predict(X_train)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = xgb_clf.predict(X_test)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1071,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.91      0.69      0.78     14018\n",
      "     default       0.41      0.77      0.54      3982\n",
      "\n",
      "    accuracy                           0.71     18000\n",
      "   macro avg       0.66      0.73      0.66     18000\n",
      "weighted avg       0.80      0.71      0.73     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1072,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.90      0.68      0.77      4673\n",
      "     default       0.39      0.72      0.51      1327\n",
      "\n",
      "    accuracy                           0.69      6000\n",
      "   macro avg       0.64      0.70      0.64      6000\n",
      "weighted avg       0.78      0.69      0.72      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1073,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7287532625592452\n",
      "0.7023428694817395\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Les résultats obtenus sont tout de suite très bons!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ce modèle est notre meilleur modèle!\n",
    "Préservons-le!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1075,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_xgb.pkl']"
      ]
     },
     "execution_count": 1075,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.externals import joblib\n",
    "\n",
    "joblib.dump(xgb_clf, \"best_xgb.pkl\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Grid Search: recall_macro (micro = bad)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Optimisons nos paramètres, pour voir jusqu'où l'on peut aller"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1076,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy as sp\n",
    "from sklearn.model_selection import RandomizedSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1077,
   "metadata": {},
   "outputs": [],
   "source": [
    "params_xgb = {\n",
    "    'max_depth': sp.stats.randint(1,10),\n",
    "    'learning_rate': sp.stats.uniform(loc=0.05, scale=0.2),\n",
    "    'n_estimators': sp.stats.randint(20,200),\n",
    "    'scale_pos_weight': sp.stats.randint(4, 20),\n",
    "    'reg_alpha': sp.stats.uniform(loc=0.01, scale=10),\n",
    "    'reg_lambda': sp.stats.uniform(loc=0.01, scale=10),\n",
    "    'random_state': [42]\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1078,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_clf = xgb.XGBClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1079,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_search = RandomizedSearchCV(estimator=xgb_clf, \n",
    "                                   param_distributions=params_xgb,\n",
    "                                   n_iter=10,\n",
    "                                   cv=3,\n",
    "                                   scoring='recall_macro',\n",
    "                                   return_train_score=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1080,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time: 82.55s\n"
     ]
    }
   ],
   "source": [
    "t0 = time.time()\n",
    "random_search.fit(X_train, np.ravel(y_train))\n",
    "t1 = time.time()\n",
    "print('Time: {:.2f}s'.format(t1-t0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1081,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['accuracy',\n",
       " 'adjusted_mutual_info_score',\n",
       " 'adjusted_rand_score',\n",
       " 'average_precision',\n",
       " 'balanced_accuracy',\n",
       " 'brier_score_loss',\n",
       " 'completeness_score',\n",
       " 'explained_variance',\n",
       " 'f1',\n",
       " 'f1_macro',\n",
       " 'f1_micro',\n",
       " 'f1_samples',\n",
       " 'f1_weighted',\n",
       " 'fowlkes_mallows_score',\n",
       " 'homogeneity_score',\n",
       " 'jaccard',\n",
       " 'jaccard_macro',\n",
       " 'jaccard_micro',\n",
       " 'jaccard_samples',\n",
       " 'jaccard_weighted',\n",
       " 'max_error',\n",
       " 'mutual_info_score',\n",
       " 'neg_log_loss',\n",
       " 'neg_mean_absolute_error',\n",
       " 'neg_mean_squared_error',\n",
       " 'neg_mean_squared_log_error',\n",
       " 'neg_median_absolute_error',\n",
       " 'normalized_mutual_info_score',\n",
       " 'precision',\n",
       " 'precision_macro',\n",
       " 'precision_micro',\n",
       " 'precision_samples',\n",
       " 'precision_weighted',\n",
       " 'r2',\n",
       " 'recall',\n",
       " 'recall_macro',\n",
       " 'recall_micro',\n",
       " 'recall_samples',\n",
       " 'recall_weighted',\n",
       " 'roc_auc',\n",
       " 'v_measure_score']"
      ]
     },
     "execution_count": 1081,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted(sklearn.metrics.SCORERS.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1082,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'learning_rate': 0.18570532016413682,\n",
       " 'max_depth': 7,\n",
       " 'n_estimators': 25,\n",
       " 'random_state': 42,\n",
       " 'reg_alpha': 1.8517598711643313,\n",
       " 'reg_lambda': 7.13948054680138,\n",
       " 'scale_pos_weight': 6}"
      ]
     },
     "execution_count": 1082,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1083,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Our hard-coded best params\n",
    "random_search_best_params = {'learning_rate': 0.09680943883172416,\n",
    " 'max_depth': 2,\n",
    " 'n_estimators': 21,\n",
    " 'random_state': 42,\n",
    " 'reg_alpha': 7.683586244347041,\n",
    " 'reg_lambda': 8.711430153536908,\n",
    " 'scale_pos_weight': 6}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1084,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6911052503346204"
      ]
     },
     "execution_count": 1084,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1085,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.718\n",
      "recall:  0.8900050226017077\n",
      "F1:  0.582703058204538\n",
      "--TEST--\n",
      "accuracy:  0.657\n",
      "recall:  0.7573474001507159\n",
      "F1:  0.49410029498525077\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = random_search.predict(X_train)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = random_search.predict(X_test)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1086,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---TRAIN---\n",
      "[[9380 4638]\n",
      " [ 438 3544]]\n",
      "\n",
      "---TEST---\n",
      "[[2937 1736]\n",
      " [ 322 1005]]\n"
     ]
    }
   ],
   "source": [
    "print('---TRAIN---')\n",
    "print(confusion_matrix(y_train, y_pred_train))\n",
    "print()\n",
    "print('---TEST---')\n",
    "print(confusion_matrix(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1087,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.96      0.67      0.79     14018\n",
      "     default       0.43      0.89      0.58      3982\n",
      "\n",
      "    accuracy                           0.72     18000\n",
      "   macro avg       0.69      0.78      0.68     18000\n",
      "weighted avg       0.84      0.72      0.74     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1088,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.90      0.63      0.74      4673\n",
      "     default       0.37      0.76      0.49      1327\n",
      "\n",
      "    accuracy                           0.66      6000\n",
      "   macro avg       0.63      0.69      0.62      6000\n",
      "weighted avg       0.78      0.66      0.69      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "# HARD-CODED\n",
    "classif_report_best_xgb = \" precision    recall  f1-score   support\n",
    "\n",
    " non-default       0.90      0.58      0.71      4673\n",
    "     default       0.35      0.78      0.48      1327\n",
    "\n",
    "    accuracy                           0.63      6000\n",
    "   macro avg       0.63      0.68      0.60      6000\n",
    "weighted avg       0.78      0.63      0.66      6000\n",
    "\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1089,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7795723500795668\n",
      "0.692925786529456\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "On sauvegarde ce modèle aussi, car il est pas mal"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1090,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_xgb2.pkl']"
      ]
     },
     "execution_count": 1090,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.externals import joblib\n",
    "\n",
    "joblib.dump(xgb_clf, \"best_xgb2.pkl\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Grid Search: recall_weighted"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Optimisons nos paramètres, pour voir jusqu'où l'on peut aller"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1029,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy as sp\n",
    "from sklearn.model_selection import RandomizedSearchCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1091,
   "metadata": {},
   "outputs": [],
   "source": [
    "params_xgb = {\n",
    "    'max_depth': sp.stats.randint(1,10),\n",
    "    'learning_rate': sp.stats.uniform(loc=0.05, scale=0.2),\n",
    "    'n_estimators': sp.stats.randint(20,200),\n",
    "    'scale_pos_weight': sp.stats.randint(4, 20),\n",
    "    'reg_alpha': sp.stats.uniform(loc=0.01, scale=10),\n",
    "    'reg_lambda': sp.stats.uniform(loc=0.01, scale=10),\n",
    "    'random_state': [42]\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1092,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_clf = xgb.XGBClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1093,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_search = RandomizedSearchCV(estimator=xgb_clf, \n",
    "                                   param_distributions=params_xgb,\n",
    "                                   n_iter=50,\n",
    "                                   cv=3,\n",
    "                                   scoring='recall_weighted',\n",
    "                                   return_train_score=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1094,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Time: 755.24s\n"
     ]
    }
   ],
   "source": [
    "t0 = time.time()\n",
    "random_search.fit(X_train, np.ravel(y_train))\n",
    "t1 = time.time()\n",
    "print('Time: {:.2f}s'.format(t1-t0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1095,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'learning_rate': 0.2338648829126646,\n",
       " 'max_depth': 8,\n",
       " 'n_estimators': 142,\n",
       " 'random_state': 42,\n",
       " 'reg_alpha': 6.31733660220611,\n",
       " 'reg_lambda': 4.992763054249679,\n",
       " 'scale_pos_weight': 5}"
      ]
     },
     "execution_count": 1095,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1096,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7629444444444444"
      ]
     },
     "execution_count": 1096,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1097,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--TRAIN--\n",
      "accuracy:  0.941\n",
      "recall:  0.9989954796584631\n",
      "F1:  0.8822355289421158\n",
      "--TEST--\n",
      "accuracy:  0.76\n",
      "recall:  0.5644310474755087\n",
      "F1:  0.5098706603131382\n"
     ]
    }
   ],
   "source": [
    "y_pred_train = random_search.predict(X_train)\n",
    "\n",
    "print('--TRAIN--')\n",
    "print('accuracy: ', accuracy_score(y_train, y_pred_train))\n",
    "print('recall: ', recall_score(y_train, y_pred_train))\n",
    "print('F1: ', f1_score(y_train, y_pred_train))\n",
    "\n",
    "\n",
    "y_pred = random_search.predict(X_test)\n",
    "\n",
    "print('--TEST--')\n",
    "print('accuracy: ', accuracy_score(y_test, y_pred))\n",
    "print('recall: ', recall_score(y_test, y_pred))\n",
    "print('F1: ', f1_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1098,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---TRAIN---\n",
      "[[12960  1058]\n",
      " [    4  3978]]\n",
      "\n",
      "---TEST---\n",
      "[[3811  862]\n",
      " [ 578  749]]\n"
     ]
    }
   ],
   "source": [
    "print('---TRAIN---')\n",
    "print(confusion_matrix(y_train, y_pred_train))\n",
    "print()\n",
    "print('---TEST---')\n",
    "print(confusion_matrix(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1099,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       1.00      0.92      0.96     14018\n",
      "     default       0.79      1.00      0.88      3982\n",
      "\n",
      "    accuracy                           0.94     18000\n",
      "   macro avg       0.89      0.96      0.92     18000\n",
      "weighted avg       0.95      0.94      0.94     18000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train, \n",
    "                            y_pred_train, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1100,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      " non-default       0.87      0.82      0.84      4673\n",
      "     default       0.46      0.56      0.51      1327\n",
      "\n",
      "    accuracy                           0.76      6000\n",
      "   macro avg       0.67      0.69      0.68      6000\n",
      "weighted avg       0.78      0.76      0.77      6000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, \n",
    "                            y_pred, \n",
    "                            target_names=['non-default', 'default']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9617605447942765\n",
      "0.6899835528411141\n"
     ]
    }
   ],
   "source": [
    "print(roc_auc_score(y_train, y_pred_train))\n",
    "print(roc_auc_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Enormément d'overfitting, la métrique choisie n'est pas optimale..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LightGBM"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### From scratch"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "List of LightGBM parameters, with their explanations\n",
    "\n",
    "https://testlightgbm.readthedocs.io/en/latest/Parameters.html\n",
    "https://github.com/microsoft/LightGBM/blob/master/docs/Parameters.rst"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 338,
   "metadata": {},
   "outputs": [],
   "source": [
    "categorical_features = ['gender', 'education', 'marriage', 'pay_1', 'pay_2', 'pay_3', 'pay_4', 'pay_5', 'pay_6']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 336,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_raw_train, X_raw_test, y_train, y_test = train_test_split(X_raw, y_raw, \n",
    "                                                           test_size=0.25, \n",
    "                                                           random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 420,
   "metadata": {},
   "outputs": [],
   "source": [
    "d_train = lgb.Dataset(X_raw_train, \n",
    "                      label=y_train, \n",
    "                      categorical_feature=categorical_features,\n",
    "                     free_raw_data=False)\n",
    "d_test = lgb.Dataset(X_raw_test, \n",
    "                     y_test, \n",
    "                     reference=d_train, \n",
    "                     categorical_feature=categorical_features,\n",
    "                    free_raw_data=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 770,
   "metadata": {},
   "outputs": [],
   "source": [
    "params = {\n",
    "    'learning_rate': 0.06,\n",
    "    'boosting_type': 'gbdt',\n",
    "    'objective': 'binary',\n",
    "    'metric': 'auc',\n",
    "    \n",
    "    \n",
    "    'scale_pos_weight': 6,\n",
    "    'num_leaves': 15,\n",
    "    'max_depth': 7,\n",
    "    'min_data_in_leaf': 500, # Min number of data needed in a child leaf,\n",
    "    \n",
    "    'bagging_freq': 1, # enabling random selection of data\n",
    "    'bagging_fraction': 0.8,\n",
    "    'feature_fraction': 0.8,\n",
    "    'verbose': 1,\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 771,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting training...\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "[50]\ttrain's auc: 0.796194\tvalid's auc: 0.773692\n",
      "[100]\ttrain's auc: 0.807604\tvalid's auc: 0.774857\n",
      "[150]\ttrain's auc: 0.817109\tvalid's auc: 0.773354\n",
      "[200]\ttrain's auc: 0.824683\tvalid's auc: 0.771904\n",
      "Early stopping, best iteration is:\n",
      "[103]\ttrain's auc: 0.808167\tvalid's auc: 0.775002\n"
     ]
    }
   ],
   "source": [
    "print('Starting training...')\n",
    "lgbm = lgb.train(params,\n",
    "                 d_train,\n",
    "                 valid_sets=[d_train, d_test],\n",
    "                 valid_names=['train', 'valid'],\n",
    "                 num_boost_round=1000,\n",
    "                 early_stopping_rounds=100,\n",
    "                verbose_eval=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 772,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving model...\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<lightgbm.basic.Booster at 0x22ea3d85828>"
      ]
     },
     "execution_count": 772,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Saving model...')\n",
    "# save model to file\n",
    "lgbm.save_model('model_lgbm.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 773,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting predicting...\n"
     ]
    }
   ],
   "source": [
    "print('Starting predicting...')\n",
    "# predict\n",
    "y_pred = lgbm.predict(X_raw_test, num_iteration=lgbm.best_iteration)\n",
    "# eval\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 774,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = np.where(y_pred>0.5, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 775,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[3394, 2479],\n",
       "       [ 331, 1296]], dtype=int64)"
      ]
     },
     "execution_count": 775,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 776,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.58      0.71      5873\n",
      "           1       0.34      0.80      0.48      1627\n",
      "\n",
      "    accuracy                           0.63      7500\n",
      "   macro avg       0.63      0.69      0.59      7500\n",
      "weighted avg       0.79      0.63      0.66      7500\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 777,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg0AAAFJCAYAAAAYOR5zAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nOzdeVjU5frH8feMiuwiiKiohYpWLqm4VFZuVLjgkqK5oBLljopiuIMo5kJxcEMpFTPXdCwxO5VWmpaVkbl0kAgPlisoIKsI8/394Y854IIgDDOM9+u6ui6Y+S73jefowzPf5/OoFEVREEIIIYR4CLWhCxBCCCFE1SCDBiGEEEKUigwahBBCCFEqMmgQQgghRKnIoEEIIYQQpSKDBiGEEEKUigwahCiHFi1a4OnpSf/+/XX/zZ0795Gvd+rUKRYsWFCBFZbOoUOHWLx4caXf9++//8bPz6/S73u3q1evMn78eBRFYdasWWzYsOG+x7zxxhsPvVaPHj04ffr0fd978803uXHjhu77c+fOMWXKFF599VV69epFnz592LRpE4Ur4TUaDW5ubvTv359+/frRu3dvxo0bx7Vr13Tvt2jRgpUrVxa7j6Io9OzZk759+wJw5swZ5s+fX7ofhhAlqG7oAoSo6jZv3oy9vX2FXCshIYGrV69WyLXKomfPnvTs2bPS73vp0iXOnz9f6fe927x58/Dz80OlUj3wGCcnJ3bs2FGu+xw7dkz3dVxcHD4+PixatEj3j/6NGzeYOHEiAD4+PgB06NCB9evX684LDg5m5cqVukFegwYN2LdvH1OmTNEdc+LECXJzc7GwsACgVatWbN26lW+//Zbu3buXqwfxeJNBgxB68tdffxEaGkpaWhoFBQV4e3szePBgtFotS5Ys4ffffycrKwtFUVi8eDENGjRg5cqVZGRkMHv2bAYMGMCiRYvYv38/AD/99JPu+1WrVnHy5EmuXbtGixYtCAsLIzIykq+++gqtVouzszNBQUE4OTnx1VdfERkZiUqlolq1arzzzjt07NixWK0ajYYvv/yS9evX4+3tTcuWLTl58iQ3btxgyJAhpKSk8PPPP5OTk8O//vUvWrRogbe3N8888wy//vorqamp9O/fX/cP18GDB1m9ejVarRYrKytmz55NmzZtitXt6urK6dOnuXr1Kr6+vmzYsIF169Zx6NAhcnNzycnJITAwkFdeeYVVq1Zx8eJFkpOTuXjxIk5OTqxYsYK6dety/vx5FixYwI0bN1Cr1UyYMIHevXtz9epVQkJCuHz5Mrdv36ZPnz6MHz/+nj+n33//nevXr9OmTZsS/zz/+ecfPD09+e2338jJySEoKIjff/8dGxsbmjVrBsDSpUsB2LlzJ0FBQdy4cYP+/fvj7+/P7NmzARg9ejRRUVH861//4q233sLd3V13D3t7e0JCQjh37tx9a7h9+zaZmZk0atRI91rz5s25fPkysbGxtG/fHoC9e/fSr18/vv/+e91xQ4cOJTg4WAYNonwUIcQja968udK3b1+lX79+uv9SUlKU27dvK71791bOnDmjKIqi3Lx5U+nVq5fy22+/KbGxsYqfn59SUFCgKIqirF+/Xhk3bpyiKIqyZ88eZezYsYqiKMrx48eVPn366O5V9PuVK1cqr732mnL79m1FURRl7969yrRp03Tf79ixQ3nrrbcURVGUnj17Kr/99puiKIry/fffK6tWrbqnj6L3HTlypDJ58mRFURTl5MmTSvPmzZVDhw4piqIooaGhyrx583THvf3220peXp6Snp6uvPbaa8o333yjJCQkKC+88IJy4cIFRVEU5YcfflC6dOmiZGRk3FN30Z7++ecfxdvbW8nJyVEURVH279+v9O3bV9dvz549lYyMDEVRFGXcuHFKRESEoiiKMmDAAOXjjz9WFEVRLl26pDvO29tbV3dubq7i7e2tfP755/f0vnTpUmXlypW67wMDA5UPP/zwnuP+/vtvpW3btoqiKEpYWJgyffp0paCgQMnIyFA8PT2VwMBARVEUpXv37kpISIiiKIpy7do1pVWrVsqlS5cURbnzv5fr168riqIobm5uSlxc3D33ufvPpX379kq/fv0UT09PpVOnTspLL72k/PPPP8X+3DZs2KAsWLBAURRFyc7OVl599VXl2LFjxf73oyiK0q5dO92fixCPQmYahCin+308kZCQwIULF5gzZ47utdzcXP744w+GDx9OrVq12LFjB3///Tc//fQTVlZWZb5v27ZtqV79zv+Fv/32W06fPs2gQYMA0Gq15OTkANCnTx8mT55M165d6dKlC2+//fZDr/3KK68A6H6jfemllwBo3LgxP//8s+64oUOHUqNGDWrUqIGHhwdHjx6lSZMmPPfcc7pzn3/+eezt7Tlz5sw9dRfl7OzM8uXLiYmJISkpSTcTU6hTp05YW1sD8Mwzz5Cenk5aWhpxcXF4eXkBUL9+fQ4ePEh2dja//PIL6enpREREAJCdnU1cXBy9e/cudt/ExMR7XnuYw4cPM3v2bNRqNdbW1gwcOLDY7EDhswSOjo7UqVOH69evU79+/WLXUBSl2MchS5Ys4aefftL92R08eBAo/vGEVqslMjKSt956iwMHDujOLXyuZu7cuXz99df06NGDatWq3VN3w4YNOX/+fLGZCiHKQgYNQuhBQUEBNjY2fPbZZ7rXUlJSsLGx4bvvviM0NBQfHx969uxJkyZN2Ldv3z3XUKlUugfi4M7UdFGWlpa6r7VaLW+99RbDhw8HIC8vj/T0dAD8/f0ZNGgQx44dQ6PRsHHjRnbv3l1i/WZmZsW+r1Gjxn2PK/qPv6IoqNVqtFrtPc8GKIpCfn7+PXUXdfbsWSZOnMiYMWPo0qULHTt2ZOHChbr3zc3NdV8X/mwK71/0fomJiTg6OqIoCjt27NB9rn/jxg1q1qx5z33v/jmXRvXq1Yudo1ar73n/Yddv164dP//8M82bNwfQDTALPwa5H7Vajbe3NytXruT69eu61x0dHXnmmWc4cuQIn376KbNmzSI1NfW+dd9vMCFEacnqCSH0wMXFBXNzc92g4fLly/Tt25czZ85w7NgxunfvzvDhw2nVqhUHDx6koKAAgGrVqun+cbW3t+fSpUtcv34dRVH4/PPPH3i/F198kd27d5OZmQlAREQE77zzDvn5+fTo0YOcnByGDRtGUFAQ586dIy8vr0L63LdvH1qtlvT0dL744gt69OjB888/z9GjR/n7778B+PHHH7l8+TLPPvvsPedXq1ZNNxj65ZdfaNWqFT4+PnTq1IlDhw7pfi4PYm1tTcuWLfn000+BOz/nYcOGkZubS9u2bdm0aRMAN2/eZNiwYRw6dOiea7i4uHDhwoUy9d21a1f27NmjmxXYv39/iQ9RFu238M93xowZrF+/nu+++043qMjNzeXrr7++ZxBS1HfffYezs/M9s1sDBgxg06ZNZGRk6AYiRSmKwqVLl3BxcSlLq0IUIzMNQuiBmZkZa9euJTQ0lA8//JD8/HymTp2Km5sbdnZ2zJgxA09PT/Lz8+nSpYvuAca2bduyZs0aJk+ezOrVq3njjTcYNGgQjo6OdOvW7YFL+by8vLh69SpDhgxBpVJRv359li5dSvXq1ZkzZw4BAQFUr14dlUrFkiVL7plJeFS5ubkMHjyYrKwshg8fzvPPPw9AUFAQkydPpqCgAHNzc9atW4eNjc095zdr1oyaNWsyePBg1q1bx1dffUWvXr3QarV0796d9PR03UDoQd577z0WLlzIli1bUKlUhIaG4ujoSFhYGIsWLcLT05O8vDz69u1Lv3797jn/tddeIzQ0tNjqg/DwcFavXq37vnv37kyfPl33/bhx4wgJCcHT0xMbGxscHByKzYQ8iIeHB97e3qxatYqnn36azZs3s2bNGt577z20Wi23bt2ic+fO7Nq1S3fOiRMn6N+/PyqVivz8fOzs7FizZs09Awt3d3eCgoLw9/e/771Pnz5N48aNadCgwUPrFOJBVEpZ5+WEEALw9vZmxIgReHh4GLqUcvP19WXq1KkPXUFR6PPPP8fa2pquXbui1Wrx8/OjS5cuuo+HjNGsWbPw8PCgW7duhi5FVGHy8YQQ4rG3cOFC1qxZU+pnG1xdXYmMjKR///707duXunXr6h7GNEZnzpxBpVLJgEGUm8w0CCGEEKJUZKZBCCGEEKUigwYhhBBClIoMGoQQQghRKrLksgR3Amm0hi6jQlWrpqKgwHQeYzG1fsD0epJ+jJ+p9ST9lE+NGg8OAJNBQwkUBdLSsg1dRoWys7M0qZ5MrR8wvZ6kH+Nnaj1JP+Xj6Hhvpkoh+XhCCCGEEKUigwYhhBBClIoMGoQQQghRKjJoEEIIIUSpyKBBCCGEEKUiMdIlULRaVCVsUSuEEEKUhTY7m+tZxbd8P3v2DJGRK1m9Oopz5+IIDPSnYcNGAAwcOJhBgwYwbtw4bt5Mp1q16tSsac57763UW40lrZ6QJZclUKnVoFIZugwhhBAmQq0okJWh+37r1s18+eUBzM0tAIiPj2Po0BEMGzay2HkXL/7Dli27UBn436RK+TVao9EQFhZW7DV/f3/y8vKYNWsWR44cue8xFSEtLY2YmJhir0VHR+vlXkIIIURZODs3JDR0he77c+f+w48/HmXSpLd5990QsrOzSElJISMjg8BAfyZM8OXYse8NVq/B5t7Dw8MxMzPT+33OnTvHN998A0Bubi4BAQFs27ZN7/cVQgghHqZbt55Ur/6/Sf+nn27JxIlTWbPmAxo0cGbjxg+4ffs2b7wxkiVLwliyZAWrVr1PauoNg9RbaR9PnDx5ktGjR5OZmYmfnx8hISF88cUXpT7/ypUrBAcHc+vWLdLS0pg0aRLu7u54enrSoUMH4uPjcXFxwcHBgRMnTmBmZkZUVBTr1q0jLi6OnTt34uHhwYABA3jhhRdITEzUY7dCCCHE/dnZWRb7PivLgurV1djZWeLp2RtbW1sA+vbtxZIloTg51WX06JFYWlpSp44tLVs+w40bV3BxaVjptVfaoMHCwoKoqChu3LiBl5cXWm3Z9nRITEzEx8eHzp07Exsby6pVq3B3dycrK4u+ffvi5uaGh4cHs2fPxt/fn5EjR5KQkMD48ePZsWMHQ4cOBeDFF19Eo9Hoo0UhhBDioe6OhL55M4f8fC1padm8/bYv/v4zeeaZVnz77RGaNm3OsWPH+OijLaxYEUF2djbnzsXj4NBAb9HSRvEgpJubGyqVCgcHB2xsbEhKSirT+Y6OjkRGRrJ7925UKhX5+fm691q2bAmAra0tTZs21X1969atimtACCGE0LOAgNmEhy+nevXqODg48M47c3F2rss33xxm7NgxqNVqxo6dhJ2dnUHqq7RBw+nTpwFITk4mOzub2rVrl+n8iIgIvLy86Nq1K3v27GHv3r2690p6mlStVpd5VkMIIYSoLPXrNyAqKhqAFi2eYt26jfccM3XqjEqu6v4q7UHI3NxcRo0axYQJEwgJCSnzshEPDw9CQ0MZPnw4P/zwA6mpqaU6r3HjxsTHxxMdHf0IVQshhBCikIQ7lUDCnYQQQlSk+4U7PYwxbY1tdOFOkydPJj09vdhr1tbWREZGVnotCipSkjMefmAVIvvMGz9T60n6MX6m1pOp9WNMZKahBDLTIIQQj+bu36iLRiX/+ec5wsNXoFarMTMzY968hdjbO7Bt2xYOHvwStVqNt7cPXbt2f6R7m9qgQWYaqgiJkRZCiEdTNC757qjkiIj38PefiatrCz79dA9bt25mzJi32b17Bzt3fkpOTg4+PsMfedAg9KfSf43WaDQsWLCA4ODgUh9/6NAhAD7++OMSj/X29uavv/4q1XUL46uFEELo191RycHBS3B1bQFAQUEBZmY1sbCwoF69+uTk5JCbm4NaZnmNkkFmGmxtbQkICCjVsa+//rru68jISEaOHFnC0UIIIYxNt249uXz5ku77OnXqAHD69O9oNLtYvfoDAOrWdcLb24uCAi3e3mMMUap4CIMMGi5evMiQIUPYtWtXqWKg69SpQ1paGunp6QQHB5c4S7Fy5UpSU1MxMzNj+fLl1KpViwULFnDlyhVSU1N5+eWXmTZtWuU1K4QQ4h6HDn3FRx9tZPnyf1G7dm2OHj3M9esp7Nq1D4AZM/xo3fpZnnmmlYErFUUZ/JmGh8VAF5owYQIff/zxQz/WePXVV+nTpw9bt25l/fr1eHt707ZtW7y8vLh165YMGoQQopIU3WOh6P4KMTH7+Oyz3Xz00UfUqnUn2bB+fUesrS2pW9cOlUpF7dq1UJTb9+zTUBrVqqkf6TxjZUz9GHzQABUbA92hQwcA2rdvz+HDh7Gzs+P06dMcP34ca2tr8vLyKq5wIYQQD1T0if/C/RWuX89gyZIlODnVY9KkyQC0a+eGr+84mjV7iiFDhqBWq2nTpi3PPNP2kVYNyOqJ8jH61ROlTYcszerQ06dP4+TkxIkTJ3B1dUWj0WBjY0NISAhJSUns2rWrVNcRQghRcYpGJX/xxTf3PcbXdxy+vuMqsSpRVkYxaCitpk2bEhAQQFhY2AOPOXjwIJs3b8bKyoply5Zx7do1pk+fzq+//oqFhQVPPPEE165dq8SqhRBCCNMg4U4lkHAnIYR4NI8Sl1xR5OOJ8jH6jyfK4tKlSwQGBt7zeseOHZkyZUqF3ktipI2fqfUDpteT9GP8TLEnoR8y01ACmWkQQhQq/M3ZFP+BNbWepJ/yMamZhsokMdJCiEJFY5Hz8vJYsmQhly5dxMrKiunTA7l8+RKRkSsxN7egc+fnGTPmLQNXLETFM/lfoy9dusQ33xR/UjcmJoahQ4caqCIhRFUXE7MXCwtLoqKi8fefyXvvLWXp0kUsXrycyMgNXLiQxO+/nzR0mUJUOJMfNBw/fpzY2Fjd9//5z3/YvXu3LLsUQjyy8+fP89xzLwDQuPGTnD79OzY2tjg7NwSgdetnOXVKBg3C9Ojl44nCTaYyMzNJTU1l0qRJKIrC1q1bdcdEREQQHR2Nk5MTI0aMID09HR8fHzQazX2vGR8fz9KlS9Fqtdy8eZN58+bRvn17XnnlFdq1a0dSUhLPPfccGRkZnDp1ChcXF5YuXUpUVBS5ubm0a9eO9u3bExYWxpw5c5g/f74+WhdCPAZcXZvzww/f8/LL3Th79gy3b9/m1q1ckpL+S8OGjfjxx2O4ujY3dJlCVDi9PdOQnZ3Npk2buHHjBl5eXgwaNIioqCgsLCxYsGABR48excvLi+nTpzNixAj279+Pp6fnA6+XkJBAYGAgLVq0ICYmBo1GQ/v27bl48SKbN2/G0dGRTp068cknnzB//nx69uxJVlYWY8eOJTExkW7duuHn58ecOXOoWbOmvtoWQpgwOztLqlVTM2LEG7z3Xhj+/hNp164dzzzTktmz5xAevgxbW1tcXZtSr15do4n+fRhjiimuCNKP/uht0NCxY0fUajV16tTB1tYWlUpFYGAgVlZWJCYm0rZtWxo1aoSVlRUJCQnExMSwdu3aB16vbt26rF27FnNzc7KysrC2tgbAzs6OBg0aAGBpaUmzZs0AsLGxKRZDffbsWZKSkggODubWrVskJCQQGhrK3Llz9fUjEEKYmLS0bOzsLDl+/AQtWrRk3LgpxMX9QWLieQ4e/IZ3330fc3Nz5syZSc+evarME/yy2sC4PRarJ86ePQtASkoKGRkZbN++ncOHDwPg4+Oje6ZgyJAhREZG4uTkhL29/QOvFxoaSlhYGE2bNmXlypVcvHgReHgEtVqtRqvV0qZNGz7//HMA/vnnH6ZPny4DBiHEI2nYsDEffLCO7ds/xtrahtmz5/PDD0eZMMGXmjVr8uqrHjRp0tTQZQpR4fQ2aEhJSWH06NFkZGQQFBSERqNh4MCBWFpaYmtrq4tydnd3JyQkhBUrVpR4vX79+jFx4kQcHByoV68eqamppaqjefPmREZG0rJlS/r06VPuvoQQws7OjoiI4jOj/foNpF+/gQaqSIjKoZdwJ41GQ2JiIgEBAQ89Nicnh5EjR/LJJ5+gNsYgJclpEEIAKArJyRkmN/UNMp1v7B6LjydKIzY2lqCgIKZNm4ZarSYvLw9fX997jnNxcSEkJKTS61O0WlSyNFMIwZ1ESCEedxIjXQKtVuH69UxDl1GhZARu/EytJ+nH+JlaT9JP+RjtTIOxU6GU+MOrqkytJ1PrB0yvp6rUT9HdGe8XFz1t2rvk5995/8KFJHr16suECX6GLFmISmPyg4ZLly4RFxdHjx49dK8tWbIEFxcXhg0bVuK5sveEEI+fontMFI2LvnDhv4SHLyc6ejNpadlcvPgPCxbMZvToez9SFcJUGeGThxWraIz0jRs3eOutt+7Zi0IIIe7n7rjo//73vO69lSvfY8IEPywtjSN0R4jK8FjFSDdv3hw/Pz+OHDmij7aFECbm7rjolJRkCgoKSEj4k6ysLDp06GToEoWoVI9NjHTPnj0BaNSokQwahBAlKozsvV9ctJlZDQ4f/po33hhqNNG+5WVMMcUVQfrRn8cmRloIIUqr8En1M2dO3RMXXVCg5dixHxg8eLjJPKEvqw2M22OxesLYYqSFEKKs7hcXDXDjxnVq1bIzcHVCVD6JkRZCiAe4X1w0wKeffmGAaoQwPImRfhhZcinE4+X/46IfxNSmvsH0epJ+ysdow50kRloIYWwkLlqIB5MY6RJIjLTxM7V+wPR6kn6Mn6n1JP2Uj9HONBg7iZGuGkytHzC9niq7n4dFQTdq1JiCggKCgmbTt+8AXYCTEKJkJj9oKBojnZCQwPz581EUhaeeeor58+dTrVq1B54rMdJCVE0Pi4KeMWMWixcHce3aVfr2HWDgaoWoOozwycOKVTRG+v3332f69Ons2LGD3NxciZMW4jFwvyjo7OxsAgPn0b59BwNXJ0TV8ljFSK9atYpq1aqRl5dHcnIyDg4O+mhfCGFE7hcF3aRJ0xJnGYUQ9/fYxUhfvHgRHx8frK2tcXFx0Vf7Qggj0adPP5KSzuPnN47WrZ+lRYunZMAgxCN67GKknZ2d+eqrr/jkk09YunQpy5Yt00P3QghDK8zqP3nyJM8/35kFC+Zz5swZkpOv6N4zM6uOtXXNMuX6G9M+ABXF1HqSfvTnsYqRHj9+PLNmzeLJJ5/EysrKOMOkhBAVonCJmp1dXcLD/8WHH27QRUEXvpeXl09m5q0yLWczteV8YHo9ST/lY5All8YYIz127FhmzZpFjRo1sLCwYPHixeXuUwhh3B4UBQ0wd25w5RYjRBUnMdIPI0suhah6HhIFXR6m9lssmF5P0k/5GG24k8RICyH0QaKghdAPiZEugcRIGz9T6wdMryfpx/iZWk/ST/kY7UyDsZMY6arB1PoB0+upMvopGh0thNAPGTSUQGKkhag6ikZH32+/ifT0dCIiwqhevRodOz7Hm2+ONXDFQlQ9Jj9oKLr3xH/+8x8WLVpEtWrVMDMzY9myZdSpU8fQJQohKtj99pu4ceMGoaHLadDAmZkzp3LuXBwtWjxl6FKFqFKMcLlCxSq690RoaCjz589ny5YtvPLKK3zwwQcGrk4IoQ937zfxn//8we3beTg7N0SlUtGp0/P8+uvPBq5SiKrnsdp74v3336du3boAFBQUULNmTX20L4QwsLv3m8jKysTZuaHufUtLSy5dumjACoWomh67vSfgzlLPjz/+uNggRghR9RVG7Y4Y8QbvvReGv/9E2rVrR4sWLbh1K1f3vqLk4+ho/0jRvMYU6VtRTK0n6Ud/Hru9Jw4cOEBkZCRRUVElxlYLIaqewmVpZ86cokWLlowbN4W4uD9ITDxPUlISZ8/G06CBM4cPH8bHZ+wjLWMzteV8YHo9ST/lY5All8a498Rnn33Gzp072bJlC3Z2duXuUQhhnBo2bMwHH6xj+/aPdftNXLlyhYUL56HVaunYsTMtW7YydJlCVDmPzd4TTz31FKGhodSvXx8/Pz/gzmzIlClTyteoEMLo3G+/iTp1HImKijZMQUKYCNl7ogSKVnsnq0EIYfQqK9zJ1Ka+wfR6kn7Kx2gTIY1+7wlUpOhp0xtDkf8zGT9T68nU+hHicSZ7T5RA9p4wfqbWD5heT9KP8TO1nqSf8jHamQZjJ3tPVA2m1g+YXk8V0U/Rjx/y8/NZvDiIK1cuo1arCQycx8aN67l+/ToAV65cpmXLVixc+G657yuE+B8ZNJRA9p4QwngU3Vvixx+PUlBQwLp1G/nll+NERa0hNPTOw9Q3b95kypTx+PnNMGS5QpikSnnKT6PREBYWVuw1f39/8vLymDVrFkeOHLnvMRUhLS2NmJgY4M4+FGPGjMHb25uRI0eSmJhY4fcTQuhfo0ZPUFBQgFarJSsri+rV//f7z8aN6xk8eIjsKyOEHhhsaUB4eDhmZmZ6v8+5c+f45ptvgDvR1SNHjmTLli2MGzeO999/X+/3F0JUPAsLC65cucTw4YNZtiyUwYPfACA19QYnTvxCr14PTpcVQjy6Svt44uTJk4wePZrMzEz8/PwICQnhiy++KPX5V65cITg4mFu3bpGWlsakSZNwd3fH09OTDh06EB8fj4uLCw4ODpw4cQIzMzOioqJYt24dcXFx7Ny5k8DAQGxs7ny2KntPCFH1FEbpRkV9wksvvYS//3QuX76Mr68Pe/d+xr///T2enn1xcNDfMyHGFOlbUUytJ+lHfypt0GBhYUFUVJRuL4rClMbSSkxMxMfHh86dOxMbG8uqVatwd3cnKyuLvn374ubmhoeHB7Nnz8bf35+RI0eSkJDA+PHj2bFjB0OHDi12rWXLlrFmzZqKblMIoUeFT5CbmVlQrVr1///ejLy829y4kcmRI0cZPdpXr0+am9qT+WB6PUk/5WMUqyfc3NxQqVQ4ODhgY2NDUlJSmc53dHQkMjKS3bt3o1KpyM/P173XsmVLAGxtbWnatKnu6/vtPXH8+HEWLlzI8uXLadKkSTk6EkIYypAhw3n33RAmTnyL27dvM3bsJCwsLLhwIYkGDZwNXZ4QJqvSBg2nT58GIDk5mezsbGrXrl2m8yMiIvDy8qJr167s2bOHvXv36t4raf+JontPHD9+nIr8MQ0AACAASURBVNDQUD788EOcneUvFiGqKktLSxYtWnrP6x9/vMsA1Qjx+Ki0QUNubi6jRo0iOzubkJAQ5s6dW6bzPTw8CA0NZf369dSvX7/Ue080btyY+Ph4oqOj0Wg03L59m1mzZgGGS5oUQgghqiJJhCyB7D0hhPGorL0lHsbUPi8H0+tJ+ikfo3imobQmT55Menp6sdesra2JjIys9Fpk7wnjZ2r9gOn1ZGr9CPE4k5mGEshMgxDFFf1t/8CBGA4cuBOclpeXR0JCPEFBi1mzJoK6dZ0A8PUdR/fuL5nUoMEUB0Gm1pP0Uz4lzTTIoOFhJEZaiP9RFJLvM/v23nvLaNbMlatXr9C8eQu6deupe0/+Ajd+ptaT9FM+JQ0a5NdoIUS5xMX9wfnzf9G//+ucOxfH55/vY+LEt1i1KrzY0mghRNVndM80lFVmZiZz584lIyOD1NRUvLy8aNWqFQsXLsTKygoHBwdq1qzJ0qVL2bJlC/v370elUtG7d29GjRpl6PKFqPI++mgTb745FoCOHTvx0kvdaNDAmRUrlvDZZ3vw9fUxcIVCiIpS5QcNSUlJ9OnTh1dffZWrV6/i7e2NlZUVy5cvx9XVlfDwcK5evUpCQgIHDhxg27ZtqFQqxowZw4svvigBT0KUQ0ZGBhcu/Jf27TsA0KdPf11U+0svdeW7774xZHlCiApW5QcNderUYfPmzXz11VdYW1uTn5/PtWvXcHV1Be4kUR44cID4+HjdLpcA6enpXLhwQQYNQpRR0Qz83347zgsvvICdnSWKojB4cF8+/ngb9erV48yZ32jX7lmjys2vCKbWD5heT9KP/lT5QcPGjRtp27Ytw4cP5/jx4xw+fJh69eqRkJBAs2bN+P333wFo0qQJzZo148MPP0SlUhEdHU3z5s0NXL0QVU/RB7L+85946tRx0r32zjvz8PObTM2a5jz5pAs+Pn0oKNDKQ2lGztR6kn7Kp0rlNJRV9+7dCQ4OJiYmBjs7O6pVq8aCBQuYM2cOlpaW1KhRAycnJ5566imef/55hg0bRl5eHm3atMHJycnQ5QtRpQ0fXvy5oE6dnqNTp+cMVI0QQt9Mcsnl1q1b6dWrF/b29oSHh1OjRg0mT578aBeTJZdC/M8DllyWRH7rM36m1pP0Uz4mPdNwPw4ODrz55ptYWlpiY2PD0qX3bmxTGopWi8r0xlRCPDJttun8RSyEKDuTHDR4eHjg4eFR7utIjLTxM7V+wDR7EkKYBpMcNFQUFUqJ0zRVlan1ZGr9gP56MpZNn4QQVZMMGkqgUqvlmQZhUtSKAln/mz3bsmUTR48e4fbt27z++mBcXZ8iLGwJNWqY4eranKlTA1DL/itCiP8nfxsI8ZiKjT3B6dOniIzcwOrVUVy9epXly0OZMmUGa9d+iJWVNV9//W9DlymEMCKVMmjQaDSEhYUVe83f35+8vDxmzZrFkSNH7ntMRUhLSyMm5s5OfMnJyYwePZrhw4czdepUcnJyKvx+QlQVP/98nKZNmzFnTgCBgf506fISycnXaN36WQBat36WU6dOGrhKIYQxMdhMQ3h4OGZmZnq/z7lz5/jmmztRtlFRUQwcOJBt27bRrFkzdu7cqff7C2Gs0tPTiIv7g0WLljFz5mwWLpxH/foN+O23XwE4dux7cnNlYC2E+J9Ke6bh5MmTjB49mszMTPz8/AgJCeGLL74o9flXrlwhODiYW7dukZaWxqRJk3B3d8fT05MOHToQHx+Pi4sLDg4OnDhxAjMzM6Kioli3bh1xcXHs3LmTOXPmoCgKWq2Wy5cv8+STT+qvYSGMVGEcbd26dWjRwhVHx1o4OtbCwsKCuXPnsGbNGnbt2kqrVq3IzMwsd3ytMUXgVgRT6wdMryfpR38qbdBgYWFBVFQUN27cwMvLC61WW6bzExMT8fHxoXPnzsTGxrJq1Src3d3Jysqib9++uLm54eHhwezZs/H392fkyJEkJCQwfvx4duzYwdChQwHIz8+nf//+3Lp1i0mTJumjVSGMWuFyzubNW/LJJ9vp338I16+nkJWVxbFjPzFz5lzq1HEkPHw5zz33QrmXf5raElJT6wdMryfpp3yMItzJzc0NlUqFg4MDNjY2JCUllel8R0dHIiMj2b17NyqVivz8fN17LVu2BMDW1pamTZvqvr5169Y916lRowYHDhzghx9+IDAwkI8//rgcXQlRdXXp8hK//x7L22+PRqvVMn16ILdv3yYgYCrm5ua0a+fG88+/aOgyhRBGpNIGDadPnwbuPIyYnZ1N7dq1y3R+REQEXl5edO3alT179rB3717de6oSlkWq1WrdrEZwcDAeHh4899xzWFlZlXieEI+DiROn3vPaiy++bIBKhBBVQaUNGnJzcxk1ahTZ2dmEhIQwd+7cMp3v4eFBaGgo69evp379+qSmppbqvMaNGxMfH090dDTe3t4EBwezZs0a1Go1wcHBJZ4rMdLC1EgMtBCiPExyw6qKotUqXL+eaegyKpR81mf8TK0n6cf4mVpP0k/5GMUzDaU1efJk0tPTi71mbW1NZGSkgSoSQgghBMhMQ4kUrfZOlLQQelZ0Twg7O0sGDhyAlZU1AA0aOOPu/hqRkSsxN7egc+fnGTPmLUOWWybyW5/xM7WepJ/yqVIzDRXt0qVLxMXF0aNHD65fv868efO4efMmBQUFLF++nMaNGz/wXNl7QlSWontCFK76Wb06CgCtVsvgwZ6sWrUeZ+eGhITM5/ffT/Lss20NVq8Q4vFk8r9GHz9+nNjYWABWrFiBp6cnW7duZdq0aSQmJhq4OiHude5cHLm5ufj7T2LKlPGcOnUSGxtbnJ0bAhLvLIQwHL3MNGg0Gg4dOkRmZiapqalMmjQJRVHYunWr7piIiAiio6NxcnJixIgRpKen4+Pjg0ajue814+PjWbp0KVqtlps3bzJv3jzat2/PK6+8Qrt27UhKSuK5554jIyODU6dO4eLiwtKlS4mKiiI3N5d27doRGxtLixYtGDNmDM7OzmVewSFEZTA3t2DYMG88PQfw998XCAiYgqIoJCX9l4YNG/Hjj8dwdW1u6DKFEI8hvX08kZ2dzaZNm3QJkIMGDSIqKgoLCwsWLFjA0aNH8fLyYvr06YwYMYL9+/fj6en5wOslJCQQGBhIixYtiImJQaPR0L59ey5evMjmzZtxdHSkU6dOfPLJJ8yfP5+ePXuSlZXF2LFjSUxMpGfPnkyZMgVbW1uio6NZvXo1H3zwAVOn3rtOXQhDKIyJtbFpQuPGjTE3N6d27aext7dn5syZhIcvw9bWFlfXptSrV9doYmUfxpgicCuCqfUDpteT9KM/ehs0dOzYEbVaTZ06dbC1tUWlUhEYGIiVlRWJiYm0bduWRo0aYWVlRUJCAjExMaxdu/aB16tbty5r167F3NycrKwsrK3vPCRmZ2dHgwYNALC0tKRZs2YA2NjY3JMIaWdnR48ePQDo0aMH4eHh+mhdiEdS+KDTv/+9jzNn/iAgYBYpKcncvHmTgwe/5d1338fc3Jw5c2bSs2evKvOglzyUZvxMrSfpp3wM8iDk2bNnAUhJSSEjI4Pt27dz+PBhAHx8fChctDFkyBAiIyNxcnLC3t7+gdcLDQ0lLCyMpk2bsnLlSi5evAiUnAYJxRMh3dzcOHz4MAMGDOCXX37RDTCEMCaDBr3O8eM/MWGCLyqVitmzF3D+fCITJvhSs2ZNXn3VgyZNmhq6TCHEY0hvg4aUlBRGjx5NRkYGQUFBaDQaBg4ciKWlJba2tly7dg0Ad3d3QkJCWLFiRYnX69evHxMnTsTBwYF69eqVOhGyefPmREZG0rJlSwIDA5k3bx47duzA2tqa9957r9x9ClHRatQwIzg4tNhrrVs/S79+Aw1UkRBC3KGXnAaNRkNiYiIBAQEPPTYnJ4eRI0fyySefoDbGTARZcikqg6KQnHxnyaVMrRo3U+sHTK8n6ad8KuTjCa1Wi6Io/Pbbb7Rp0wYzM7NyFxYbG0tQUBDTpk1DrVaTl5eHr6/vPce5uLgQEhJS7vuVlew9ISqL7AkhhKgKSjXTsGLFCho1asSlS5c4e/YsderUYdmyZZVRn0HJ3hPGz9T6AdPrSfoxfqbWk/RTPuWeafj111+ZOXMm3t7ebNmyhdGjR1dYccZMhVLiD6+qMrWejLWfotHQhXx8hheLh+7S5SXWrImgbl0nAHx9x9G9+0uVXqsQQpRGqQYNWq2WU6dO0bBhQ/Ly8rhx44a+66owRWOkz549y/jx43nyyScBGDZsGL17937guRIjLcqjaDQ03BsPDRAVtZaJE6fQrVvPSq9PCCHKqlSDhv79+7No0SKWLFnCihUrGDVqlL7rqjDHjx8nMTGRHj168Mcff+Dj48Obb75p6LLEYygh4U9dPHRBQQFjx07i3Lk4/vzzHLt2befpp1syYYKfocsUQogHKvXqiYyMDC5dukSjRo2wtCw5mcqYYqT79OlDbm4u8+fP58iRI5w/f56CggKeeOIJ5syZowuJevBPSGYaxCMqsiIC4K+/Ejh79nSxeOj+/V+nW7eeNGjgzIoVS2jatBm+vj7yeawRM7V+wPR6kn7Kp9zPNHz55ZdERkZSUFCAh4cHKpWKiRMnlniOMcZIp6Wl4eXlRatWrYiMjGTNmjUEBgaW5kcgRLk1atSYhg0bolKpaNz4CWrVqoW7+2s4OdUD4KWXuvLdd98YuEohhHiwUg0aNm3axK5du/D19WXixIkMGjTooYMGY4yRfuWVV7C1tdV9vWjRotK0L8QjK5oXv2PHPv78M5758xdw7do1cnKyGT/eh+3bd1KvXj3OnPmNdu2eNaqc+Yog/Rg/U+tJ+tGfUg0a1Go1ZmZmqFQqVCoVFhYWDz3HGGOkfX19mT9/Pm3atOHHH3+kZcuWD29eiHIoOqXYs2cvjh//iWHDhv3/IHo+OTk5+PlNpmZNc5580gUfnz4UFGhlatWImVo/YHo9ST/lU+6PJzp06MCMGTO4evUqCxYsoHXr1g89xxhjpIODg1m0aBE1atSgTp06MtMgKlWNGjXuiYcG6NTpOQNUI4QQZVeqByEzMjL47bffiI+Pp0mTJrqdIh9EYqSF4J4HIUtLfksybqbWD5heT9JP+ZR7pmHs2LFs376dl19+ucKKAomRFqZNoqGFEKamVDMN48eP5/nnn8fFxUU3G/Diiy/qvThDkxhp42dq/YDp9ST9GD9T60n6KZ9yzzTUrl2buLg44uLidK89DoMGiZGuGoy1n/vFSAshRFVWqkHDu+++q+86jJLESIvyuDtGGmTvCSFE1VaqQUPRWYW0tDQaNWrEF198obeihDBFsveEEKKqK9VyhaNHj+r++/LLL2nbtm2ZbqLRaAgLCyv2mr+/P3l5ecyaNYsjR47c95iKkJaWRkxMDHAnpfKdd95h+PDheHl5cerUqQq/nxAPUnTviSlTxnPmzGnOnYvj88/3MXHiW6xaFU5+fr6hyxRCiAcq8xpHZ2dnEhMTy33j8PBwzMzMyn2dhzl37hzffHMnmnfDhg24urqybds2Fi1aVCF9CFFa5ubmDBvmzfvvryYgYDYhIfNo396NadNmsmbNB+TkZPPZZ3sMXaYQQjxQqT6emD59ui558dq1azg4OJT5RidPnmT06NFkZmbi5+dHSEhImT7iuHLlCsHBwdy6dYu0tDQmTZqEu7s7np6edOjQgfj4eFxcXHBwcODEiROYmZkRFRXFunXriIuLY+fOnRw9epRevXrh6+uLlZUVQUFBZe5DiLIoGv3auvVTtGzZHHNzc2rXfhp7e3tef30A9evXB8DD41W+/vpro4qMrQjSj/EztZ6kH/0p1aDhjTfe0H1ds2ZNWrVqVeYbWVhYEBUVpdvAqjDaubQSExPx8fGhc+fOxMbGsmrVKtzd3cnKyqJv3764ubnh4eHB7Nmz8ff3Z+TIkSQkJDB+/Hh27NjB0KFD2bBhAzdv3mTDhg18+umnLFu2jOXLl5e5FyFKq+gyqb17d/PXXwkEBMwiJSWZ9PR03nhjKB98sJm6dZ04fPgoLi6uEiNt5EytHzC9nqSf8nnkJZcFBQUUFBTw0UcfER4ejqIoKIqCj48PH330UZmKcHNzQ6VS4eDggI2NDUlJSWU639HRkcjISHbv3o1KpSr22W/hHhK2trY0bdpU9/XdG1bZ2dnp0iy7d+9OVFQUQlSWvn37ExoazIQJvqhUKubMCSInJ4e5c2fq9p7o12+gocsUQogHKnHQsGfPHtatW0dKSgoeHh4oikK1atVwc3Mr841Onz4NQHJyMtnZ2dSuXbtM50dERODl5UXXrl3Zs2cPe/fu1b1X0qZVRTescnNz4/Dhw7Rq1YpffvlFtyOmEJVB9p4QQlR1JQ4ahgwZwpAhQ9i9ezeDBw8u141yc3MZNWoU2dnZhISEMHfu3DKd7+HhQWhoKOvXr6d+/fql3rCqcePGxMfHEx0dzbhx45g3bx5Dhw6levXqLFu2rMRzJUZalIfESAshTE2pYqSTkpL497//ze3bt4E7D0MaYi+IyiYx0sbP1PoB0+tJ+jF+ptaT9FM+5Y6RDgwMpHv37sTGxlK3bl2y9fgb1OTJk0lPTy/2mrW1NZGRkXq7pxBCCCEerlSDBnNzc8aNG8d///tf3n33XYYPH663glavXq23a5eV7D1RNRiyn7v3l7g7JvrVV3vxwQeRVK9endq1azNvXgjm5uaGKlcIIcqlVIMGRVFITk4mKyuL7Ozse2YCqpro6GhSUlIICAgo8TjZe0I8TNH9Je4XEz1s2OusWfMB9vYOrFu3mpiYT/HyeuO+1xJCCGNXqkTIyZMn8/XXX9O/f3969uzJyy+/rO+69CI3N5eAgAC2bdtm6FKECbpfTPSqVVHY298JQysoKKiUFFQhhNCXUs00dOzYkaeffpqLFy9y8OBBrKys9F0XGo2GQ4cOkZmZSWpqKpMmTUJRFLZu3ao7JiIigujoaJycnBgxYgTp6en4+Pig0Wjue81bt24xYMAAXnjhBYmQFhWuMCba03MAf/99gYCAKWzbdicW+vDhb4mNPcFbb403cJVCCPHoSjVo+PLLL4mMjKSgoAAPDw9UKhUTJ07Ud21kZ2ezadMmXYrkoEGDiIqKwsLCggULFnD06FG8vLyYPn06I0aMYP/+/Xh6ej7werVq1eLFF1984KBCiEdRGO96v5jo27ezOHDgK7766is+/PDDUuWTGFNkbEWQfoyfqfUk/ehPqQYNmzZtYteuXfj6+jJx4kQGDRpUKYOGjh07olarqVOnDra2tqhUKgIDA7GysiIxMZG2bdvSqFEjrKysSEhIICYmhrVr1+q9LiGKKlwKdXdM9M2bN9m+fScJCX/y3nurUKlqlmrZlCwXM26m1g+YXk/ST/mUe8mlWq3GzMwMlUqFSqXCwsKiwoorydmzZwFISUkhIyOD7du3c/jwYQB8fHwojJgYMmQIkZGRODk5YW9vXym1CXG3u2OiZ86cg7//JJo3f4oZM6YA0LPnqwwcWL6gNCGEMJRSDRo6dOjA9OnTuXr1KgsWLKB169b6rgu4M1gYPXo0GRkZBAUFodFoGDhwIJaWltja2nLt2jUA3N3dCQkJYcWKFZVSlxD3c7+Y6O++O26gaoQQouKVOGhYu3YtEydOZPr06cTExPDMM8/QpEkT3aZP+taxY8diyyK7du163+MKCgpwdnamS5cupbru66+/XiH1CSGEEI+TEgcNx48f1z278Mknn5R5Z8vKEBsbS1BQENOmTUOtVpOXl4evr+89x7m4uJQ5+lr2nhAPI/tLCCEeJyUOGopuS1GKLSoqVGlnA9q3b09MTIzuezMzM7Zs2VIhNSioSEnOqJBrGQt5QEgIIcSjKnHQUHTL6ZK2nzZVEiNdNeizn7tjolNTb+Dr6014+Bpyc3MJC1tCjRpmuLo2Z+rUANTqUuWlCSFElVTioOHs2bO88cYbKIpCQkKC7muVSsWOHTsqq0aDkRhpUTQmOj8/n+XLl2BmVhOA5ctDmTYtgNatnyUqai1ff/1vXnuttyHLFUIIvSpx0LBv374KuYlGoyExMbHYQ43+/v4sW7aMBQsW0Lt3b1JSUu45piKkpaXx/fffFwt9+uWXXwgICNAt3xSiNFav/hcDBgxiy5ZNACQnX6N162cBaN36WY4ePSyDBiGESStxLtXZ2fmB/5VXeHh4peTwnzt3jm+++Ub3/eXLl9m4cSP5+fl6v7cwHQcOxGBnZ0fnzs/rXmvQwJnffvsVgGPHvic3N8dQ5QkhRKUoVU5DRTh58iSjR48mMzMTPz8/QkJC+OKLL0p9/pUrVwgODubWrVukpaUxadIk3N3d8fT0pEOHDsTHx+Pi4oKDgwMnTpzAzMyMqKgo1q1bR1xcHDt37mTAgAEEBQWxaNEiWXYpyuTzz/ehUqk4ceJnEhLiWbx4ARMnTmXLlmi2bfuIp556BjOzGoYuUwgh9KrSBg0WFhZERUXp9pHQarVlOj8xMREfHx86d+5MbGwsq1atwt3dnaysLPr27YubmxseHh7Mnj0bf39/Ro4cSUJCAuPHj2fHjh0MHTqUuXPn8uabb+Lk5KSnLoUpsrOzLLZR2pgxo1mwIIjvvz/C0qXvUrduXZYsWUy3bi9XSD68MeXMVwTpx/iZWk/Sj/5U2qDBzc0NlUqFg4MDNjY2JCUllel8R0dHIiMj2b17NyqVqtjHCy1btgTA1taWpk2b6r6+deuW7pirV69y4sQJLly4wJo1a0hPT8ff35/w8PAK6E6YsruXdObnF5CRkYuDQz3Gjh2Lubk57dq50aZNxwpZ/mlqy0ilH+Nnaj1JP+VT7r0nKsLp06cBSE5OJjs7u1S7/RUVERGBl5cXXbt2Zc+ePezdu1f3XknLQdVqNVqtFicnJ7788kvd6126dJEBg3gkq1dHAfDEE0/y4osvG7gaIYSoPJW2qDw3N5dRo0YxYcIEQkJCypz74OHhQWhoKMOHD+eHH34gNTW1VOc1btyY+Ph4oqOjH6FqIYQQQhRSKZUd9ViFKFrtnawG8di6O9ypMsjUqnEztX7A9HqSfsrHKD6eKK3JkyeTnp5e7DVra2siIyMrvRaJkTZ+ptaPEEIYM6MbNKxevdrQJehIjHTVUFH9GGJWQQghqhKjGzQYE4mRfrwUjYyG4vtMbNy4nuvXrwNw5cplWrZsxcKF7xqqVCGEMIhK+cBeo9EQFhZW7DV/f3/y8vKYNWsWR44cue8xFSEtLU23C2ZaWhqdO3fG29sbb29vNm/eXOH3E6bh7n0mFi58l9Wro1iyJAxraxv8/GYYuEIhhKh8BptpqKzljoUx0p6envzxxx/07duX+fPnV8q9RdV19z4ThTZuXM/gwUOoU6eOgSoTQgjDeaxipNPT0zl79iwjR47E3t6eefPmUbduXT12LaqiovtMFB00pKbe4MSJX/Dzm27A6oQQwnAeqxjpgwcP0qpVK1544QX27dvH4sWLWblypZ46FlWRnZ0lX365H5VKxbRpv/LXX3/y7rsLWb16NT/99D2enn1xcNDvg6TGFBlbEaQf42dqPUk/+vPYxEgDPPfcc1hYWADwyiuvyIBB3CMtLZuIiPW67ydPHsvMmXOoXt2KI0eOMnq0r96XeJraMlLpx/iZWk/ST/mUtCKt0pKLKiJGun///qxYsYLOnTtTNJOqNDHSAPPmzdNFSf/444+6wYYQpXHhQhINGpR/W3ghhKiqKm2moTBGOjs7m5CQEObOnVum8wtjpNevX0/9+vUfKUZ6xowZzJkzh+3bt2NhYcHixYsfpRXxGCncZwLg4493GbASIYQwPImRLoHESD9ejCXcSaZWjZup9QOm15P0Uz4SI/2IJEba+JlaP0IIYcyMbtBgTDHSQgghhPgfoxs0GBPZe8Lwin5kUFBQwLJli/n77yTU6mrMmRPE4sXruXLlKiDxzkIIoW+P1aAhOzub4OBg/vnnH27fvs38+fNp06bNA4+XvScMr+h+EMeOfQ9AZORGYmNPsGrV+6xbt460tGxu3rzJlCnjJd5ZCCH06LEaNGzYsAFXV1eWL19OXFwccXFxJQ4ahHF5+eVuvPDCiwBcvXqF2rUddO9JvLMQQuif0Q4aNBoNhw4dIjMzk9TUVCZNmoSiKGzdulV3TEREBNHR0Tg5OTFixAjS09Px8fFBo9Hc95pHjx6lV69e+Pr6YmVlRVBQUGW1IypI9erVWbw4iCNHvmPx4mWAxDsLIURlMdpBA9z5OGHTpk266OlBgwYRFRWFhYUFCxYs4OjRo3h5eTF9+nRGjBjB/v378fT0fOD1UlNTuXnzJhs2bODTTz9l2bJlLF++vBI7Eo/i7vjUsLAVpKQkM2zYG3Tt+kKlxTtXFmOKjK0I0o/xM7WepB/9MepBQ8eOHVGr1dSpUwdbW1tUKhWBgYFYWVmRmJhI27ZtadSoEVZWViQkJBATE8PatWsfeD07Ozt69OgBQPfu3YmKinrgscJ4FC6p/Pe/Pyc5+Rre3j7cuqUAKkBVafHOlcXUlpFKP8bP1HqSfsqnSuU0FHX27FkAUlJSyMjIYPv27Rw+fBgAHx8fXZT0kCFDiIyMxMnJCXt7+wdez83NjcOHD9OqVSt++eUXmjVrpv8mRIXp2rUHS5YsZNKkt8nPz2fKlOnUrFlT4p2FEKKSGG0ipEajYefOnZibm5ORkcHUqVPRaDScP38eS0tLbG1tadeuHRMmTCAvL4+XX36ZFStW8NJLLz3wmmlpacybN4/k5GSqV6/OsmXLaNiwYcmFyOoJw1IUkksI2DK13yjA9HqSfoyfqfUk/ZRPlZ1p6NixIwEBAbrvu3btet/jCgoKcHZ2pkuXLiVez87OTsKjhBBCiEdk1IOG0oiNjSUoKIhp06ahVqvJy8vD19f3nuNcXFwICQkp07UVrRaVcU7EPDa02abzvqj9qQAAIABJREFU24IQQlR1RvvxhDHQahWuX880dBkVSqbtjJ+p9ST9GD9T60n6KZ8q+/GEoUmMdMV6WCR0dnYW4eErUKvVmJmZMW/eQuztHR5yVSGEEJXlsRo0XLp0iTlz5lBQUICiKISEhNCkSZMHHi8x0hXrYZHQmZmZ+PvPxNW1BZ9+uoetWzdLYJMQQhiRx2rQEBERwciRI3F3d+f777/n/ffflwcjDeR+kdABAXN0MdAFBQWYmdU0ZIlCCCHuYrSDBn3ESAcGBmJjc2dqvqCggJo15R8lQ7o7ErpwwHD69O9oNLtYvfoDA1cohBCiKLWhCyhJYYz0xo0bWbp0KQkJCURFRbFlyxZcXFx0MdKffvopwENjpO3t7alRowaJiYksW7aMSZMmVVYr4gHmzVvI9u17WLZsMTk5ORw69BVhYe+yfPm/qF27tqHLE0IIUYTRzjRAxcdIAxw/fpyFCxeyfPnyEp9nEPpRmJ++b98+rl69wttvj6V6dXuqV6/GiRPH+Oyz3Xz00UfUqmVXqusZUyZ7RTG1nqQf42dqPUk/+mPUg4aKjpE+fvw4oaGhfPjhhzg7S+ywIRQuG+rYsQtLlixkxIgR5OfnM3myP0uWhODkVI9JkyYD0K6dG76+40q8nqktrQLT60n6MX6m1pP0Uz4lrbAz2pwGfcRI9+vXj7y8PBwdHYFSBj7J6omK85BI6Edhan85gOn1JP0YP1PrSfopnyqb01DRMdL79u2r0PqEEEKIx4lRDxpKQ2Kkqw6JhBZCiKrNaD+eMAYSI238TK0fML2epB/jZ2o9ST/lU2U/njA0iZGuGEXjo4UQQlRdMmgogcRIV4yi8dH323PC2bkhACtXvkfjxk8wYMBgQ5YrhBDiAYw63OlBoqKiOHXqlKHLEI+g6J4Tvr7jWLXqfVJTU5kxYwpHjx4xcHVCCCFKUiVnGsaOHWvoEsQjut+eEzk52bz55liOHz9m4OqEEEKUpFIGDRqNhm+//Zbc3FySk5MZNWoUhw4d4s8//+Sdd97hypUrfPXVV+Tn52NjY8OqVavYv38/e/bsQavVMmXKFObMmUOTJk1o0qQJGRkZ9O7dm/bt2zN37lwyMjJITU3Fy8uL4cOHc+rUKRYuXIiVlRUODg7UrFmTpUuXsmXLFvbv349KpaJ3796MGjWqMtoXd7l7z4kGDZxp0MBZBg1CCGHkKm2mISsri40bN/L5558THR3Nrl27+Omnn4iOjqZVq1ZER0ejVqvx9fXl9OnTANja2hIZGQnA5cuX0Wg01K5dm1mzZgGQlJREnz59ePXVV7l69Sre3t4MHz6coKAgli9fjqurK+Hh4Vy9epWEhAQOHDjAtm3bUKlUjBkzhhdffFGipCvJ3RGoYWErSElJZtiwN/jssxgsLS0xN6+BhYVZmeJSjSletaKYWk/Sj/EztZ6kH/2ptEHD008/DYCNjQ1NmzZFpVJRq1Ytbt++TY0aNZg+fTqW/9fenUdFXbf/H38Oi6yiojhqWrJEWt3eitptLrhkuRw7FYkSRKLeluRO5PhDEFRAvSHMJUgzxdu81bybc1LzTtM6Kpi5tohaiqmlJiKIjCjbfH5/8HVCA0UF5jPD9TjHcxgY3lwvEbzmM/O+3s7O/PHHH5SVlQEVsxVuadas2V8OMGrRogWrV69m+/btuLq6mj4vJyeHxx9/HICuXbuydetWfvnlFy5cuEBYWBgABQUFnDt3TpqGenJru9CXX37B5cs5hIaOprhYATQUFhZTUgI3b5Zy40bJfW0tsratVWB9mSSP+llbJsnzcFSx5VJTzS6E0tJSduzYwcaNG7lx4wYBAQGmMyVsbP58nWblt29ZuXIlnTt3Jjg4mH379pnOpWjVqhWnTp3Cx8eHH374AQAvLy98fHxYsWIFGo2G9PR0fH19azumuIe+fQeQmDibCRPGUVZWxuTJEXJEuRBCWAizvxDSzs4OJycnAgICaNSoER4eHuTk5NToc/v3709cXBybN2+madOm2NraUlJSQmxsLFFRUTg7O2Nvb49Wq6VDhw48++yzvPbaa5SUlNCpUye0Wm0dpxN3cnJyYu7c+VV+7F6HUwkhhDAvq5wIuXbtWoYMGYK7uzsLFy7E3t6eiRMn3vc6itFYMatBPJS6HO5kbZchwfoySR71s7ZMkufhqOLpifrUvHlzxowZg7OzM40bN2b+/Kof2d6LgobcWj6V0dys7YdJCCFE/bHKpmHw4MEMHjzY3GUIIYQQVsUqm4baImdPPLjKT0lUNTpaURQSEuLQaDR4eXkTEaGr8sWuQggh1KNBNQ2XL18mMjKS0tJSPDw8mD9/Pk5OTtXeX86eeHCVz5uoPDr68OGDLFmSgqIojBsXjp9fN5KSEtmzZxd9+/Y3Z8lCCCHuoUE9tFu+fDmvvPIK//nPf/Dx8WHDhg3mLqlB8Pfvx/TpM4E/R0f//PMJunTpCkCPHj05eHC/OUsUQghRA6q90qDX69m5cycGg4H8/HwmTJiAoiisXbvWdJ9FixaRnp6OVqslJCSEgoICRo8ejV6vr3LNqKgoFEXBaDRy8eJF2rdvX09pxJ2jo/fu3WOa3eHs7ML16wYzVyiEEOJeVNs0ABQVFbFq1Sry8vIIDAzk1VdfZfny5Tg5OTFr1iwyMjIIDAwkIiKCkJAQtmzZwosvvljtehqNhrKyMl566SWKi4uZMGFCPaZpeO42Orq0tMT0cY2mjObNmz3QmFQ1jVetLdaWSfKon7Vlkjx1R9VNQ/fu3bGxsaFFixa4ubmh0WjQ6XS4uLhw+vRpOnfuTLt27XBxceHUqVNs3ryZ1NTUu65pb2/P1q1b2bt3Lzqdjk8++aSe0jQ8dxsd7evbga+/3o2fXzd27vwGP79uD7QV1Bq3kFpbJsmjftaWSfI8HIud05CVlQVAbm4uhYWFrFu3zjQqevTo0aZx0yNGjCAtLQ2tVou7u3u168XFxTF48GB69OiBi4tLtaOtRe2qanT0Y4958q9/JbBs2Qc89lh7+vV7ztxlCiGEuAfVToTU6/Vs2LABR0dHCgsLmTJlCnq9nl9//RVnZ2fc3Nzo0qUL4eHhlJSU4O/vT1JSEn369Kl2zezsbOLi4oCKsyxmzZqFt7f33QuRxuLBKAqX62EwlrU9ogDryyR51M/aMkmeh2OxVxq6d+9OZGSk6Xbfvn2rvF95eTmPPPIIvXr1uut63t7erFmzplZrFEIIIRoKVTcNNXH48GFiY2OZOnUqNjY2lJSUMHbs2L/cz9PTkzlz5tzX2orRiEadF2JUz1hkPV2+EEKICqp9ekINjEaFK1esayugXLZTP2vLJHnUz9oySZ6HY7FPT5ibjJGuUHkkdFlZGfPmzebixYuUlpYwatRYPDy06HTTaNu2HQCvvDKc5557odbrFkIIYV5W2TTs3r2brVu3PvDplrfIGOkKlUdCb9u2FTe3psTEzKWg4CqjR4cwevQ4Ro4M4bXXXjdzpUIIIeqSVTYNou707z+Q/v3/3B5pa2vHzz8f59y5s2Rk7KJt23ZMmfIOzs4uZqxSCCFEXVBN03Dz5k2mT59OTk4OrVu35sCBA3z88cfEx8cD0LRpUxITEzl27BgfffQR9vb2/P777wwdOpTw8HCys7OJiorCyckJJycnmjRpAsD//vc/0tPTsbGxoWvXrkRGRrJkyRKOHDlCUVERCQkJ9952KUycnSumkhUVXSc6Wse4ceGUlpYwbNjLdOjQkdWrP2blyo+YOHGqmSsVQghR21RzYNWGDRto27Yt69evZ+LEiVy5coWYmBhiY2NZs2YN/v7+rFixAoALFy6wZMkSNmzYYHrfokWLmDx5Munp6XTp0gWAq1evsmTJEtLT01m3bh2XLl0iMzMTAC8vL9avXy8NwwO4dOkPJk0az6BBQ3nhhcH4+/enQ4eOAPj79+fkyZ/NXKEQQoi6oJorDdnZ2fj7+wMV8xTc3d3Jzs5m9uzZAJSWluLp6QmAr68vdnZ22NnZ4ejoCMDJkyfp1KkTAH5+fpw+fZpz586Rl5fHm2++CcD169f57bffAExriZq5Nfc8NzeXyMjJzJw5kx49ngUgPHw0UVEz+dvfOnHs2Pd06vS3epuTrqaZ7LXF2jJJHvWztkySp+6opmnw9fXlyJEjDBw4kHPnzpGfn8+TTz7JggULaNOmDYcOHeLy5csAVY5/9vLy4siRI/j7+3P06FEA2rZtS+vWrVm5ciX29vbo9Xo6duzIjh07sLFRzUUWi3Bru8/SpakUFBSwdOkHLF36AQDh4VNISEjEzs6O5s2bM336zHrbHmRtW6vA+jJJHvWztkyS5+FYxJbL4cOHM2PGDEJCQmjTpg0ODg7ExcWh0+koL6/Y7peQkEBOTk6Vnx8bG8u0adP4+OOPcXd3x8HBAXd3d8LCwggNDTVNjRwyZEh9xrI6U6dGMnVq5F/e/+GHK81QjRBCiPqkmuFOhw8fpqioiN69e3PmzBn++c9/smPHDnOXJVsuod7OkXgQ1vaIAqwvk+RRP2vLJHkejkVcaWjXrh0REREsXbqUsrIyZs2aZe6SZIz0/5GR0EIIIUBFTYOHh4fqDpNS0JCr0kfYD8raOnAhhBD1RzVNgxo15DHSlUdHCyGEECBNw1015DHSlUdHV3XeRO/eFceUb9/+JZ99toFly1aZs1whhBD1wCz7DpOTk9Hr9bWy1ldffcWlS5e4fPkycXFxtbKmuN2t8yZSU1eQnLyYlJR/AXDy5M988cXnqOS1tEIIIeqYxQ8r+Pe//43BYMDDw0OahjrSv/9Axo0bb7pta2tHQcFV0tKWMnnyO2asTAghRH2qk6cnSktLiY2N5ezZsxiNRqZOncrVq1dJS0vD3d2d0tJSvLy8+O6771i/fj0LFy4EoFevXmRmZnLmzBmio6MpLS3F0dGRhQsXkpuby/z58zEajVy7do3o6GiuXbvG8ePH0el0JCUlodPp+PTTT8nMzOT999/HwcHBdGbF8ePHqzyzQtzbX8+bGM/8+XOZPDkCBwcHM1cnhBCivtRJ07Bx40aaNWtGYmIi+fn5vP766xQXF7Nx40aaNm1qGutcnQULFvDmm2/i7+/P1q1bOXbsGNeuXUOn0/HEE0+wefNm9Ho98fHxdOzYkbi4OOzt7QFQFIWYmBjWrVuHVqtl9erVpKWl0a9fPy5cuMCmTZsoKSmhT58+0jTcQ+WxpRcvXmTq1MkEBQXx+OOPs3ZtOu+//y9KSoo5e/ZXPvxwETNm/L96r1FN41Vri7VlkjzqZ22ZJE/dqZOm4ZdffuHQoUP8+OOPANy4cQOAZs2aAZgOlLrTrefGf/31V9N9hg4dCsDBgwdJTU3F0dGR69ev4+rqWuUa+fn5uLq6otVqAejevTspKSn069evyjMrRPVubc3My7vCpElvMW3adLp1ewaA1as3AHDx4gViY6MYP36KWbZyWuMWUmvLJHnUz9oySZ6HU+/Dnby8vGjVqhXjx4/n5s2bpKWlsWXLFvLy8nB3d+enn36iVatWODg4mM6TOH/+PAUFBUDFgVU//fQTPXv2ZNOmTRQUFKDX60lOTsbb25vFixdz/vx5oOIcisovxGvWrBkGg4GcnBxatmzJ/v37ad++vem+4v79+9+rKCwsJD19BenpFaeKvvfeYhwcpPESQoiGpE6ahqCgIKKjo3n99dcxGAwEBwczb948xo4dS5MmTbCzq/iyTz/9NI0bNyYwMBBvb2/atm0LwPTp05k1axZpaWk4OjqSlJREWVkZb7/9Ns2bN6dVq1bk5+cDFVctpk+fzty5c4GKxiA+Pp5Jkyah0Who0qQJ8+bN4+TJk3URtUGo7rwJgNat27B8eXr9FiSEEMIsVHP2hBopRmPFrIYGyFKGO1nbZUiwvkySR/2sLZPkeTgWcfaEGskYaSGEEOJPDfNhtBBCCCHum1xpuAtrPnui8tMPdxsTvXjxezz66GO8/PJwc5YshBBCBRrklYYDBw7Qt2/fe97PdPaEFf6xcf5zz29VY6Lz8/N5553JZGTsrstvhRBCCAvS4K40XLx4kZUrV1JWVmbuUlSjf/+B9O//nOm2ra0dN24UMWbMm+zbl2nGyoQQQqiJapsGvV7Pzp07MRgM5OfnM2HCBBRFYe3atab7LFq0iPT0dLRaLSEhIRQUFDB69OhqD8MqLi4mNjaWuXPnEhAQUF9RVO+vY6LDadPmEdq0eUSaBiGEECaqbRoAioqKWLVqFXl5eQQGBvLqq6+yfPlynJycmDVrFhkZGQQGBhIREUFISAhbtmzhxRdfrHa9OXPmMGbMGNO0yIauujHRlRsqR0d7nJwaqWaE6Z3UNF61tlhbJsmjftaWSfLUHVU3Dd27d8fGxoYWLVrg5uaGRqNBp9Ph4uLC6dOn6dy5M+3atcPFxYVTp06xefNmUlNTq1zr0qVLHDx4kHPnzvHBBx9QUFDAtGnTTIdlNUTVjYmuvCXz5s1SbtwoUe02TWvcQmptmSSP+llbJsnzcCx2TkNWVhYAubm5FBYWsm7dOnbt2gXA6NGjTeOjR4wYQVpaGlqtFnd39yrX0mq1bNu2zXS7V69eDbphqEzGRAshhKgJVTcNubm5jBo1isLCQmJjY9Hr9bzyyis4Ozvj5uZGTk4OAAMHDmTOnDkkJSWZuWLLdLcx0WPHvlXP1QghhFArVTcN3bt3JzLyz//MqtsmWV5eziOPPEKvXr1qvHZmprzATwghhLgfqm4aauLw4cPExsYydepUbGxsKCkpYezYsX+5n6enJ3PmzLmvtRWjEY2VHs1hLLKe5/uEEELUDzmw6i6MRoUrVwzmLqNWyQuE1M/aMkke9bO2TJLn4VjsCyHNzRrHSCs3bpjermp8dPv2XiQkxKHRaPDy8iYiQodNAz3pUwghxO0aVNNw9epVBg0ahK+vL1DxAspRo0ZVe3/TGGkrolEUMFSc3HlrfHRMzFwKCq4yenQIjz/uy7hx4fj5dSMpKZE9e3bRt29/M1cthBBCDRpU03Ds2DGGDRtGTEyMuUtRharGR//88wm6dOkKQI8ePdm//ztpGoQQQgAqbhrqYoz00aNHycrK4vXXX8fd3Z3o6GhatmxZX5FUp6rx0R988D6a/7u64uzswvXr1vWaDiGEEA9OtU0D1P4YaS8vL55++ml69uzJpk2biI+PZ/HixfWYSB3uNj562bIlpo9rNGU0b95MNeNLq6Km8aq1xdoySR71s7ZMkqfuqLppqM0x0gA9evTAyckJgOeff75BNgxw9/HR3t6+fP31bvz8urFz5zf4+XVT9auQre1V0mB9mSSP+llbJsnzcCx290RtjpEGiI6O5oUXXmDo0KF8++23PPXUU3UfQsWqGh89ZUokixYls2zZBzz2WHv69XvuHqsIIYRoKFQ7p0Gv17NhwwYcHR0pLCxkypQp6PV6fv31V9MY6S5duhAeHk5JSQn+/v4kJSXRp0+fatf87bffiIqKAsDJyYn4+Ph7v6bBynZPoChcvlxo7ipqjbU9ogDryyR51M/aMkmeh2OxVxpqe4x0u3btWLNmTa3WKIQQQjQUqm4aakLGSN+fysOdhBBCiPuh2qcn1EDGSKufteUB68skedTP2jJJnodzt6cnZD6wEEIIIWpEmgYhhBBC1Ig0DUIIIYSoEWkahBBCCFEj0jQIIYQQokZk94QQQgghakSuNAghhBCiRqRpEEIIIUSNSNMghBBCiBqRpkEIIYQQNSJNgxBCCCFqRJoGIYQQQtSINA1VMBqNzJo1i5EjRxIaGsrZs2fNXdIDKS0t5d133yU4OJjhw4ezc+dOzp49y2uvvUZwcDCxsbEYjUZzl3nfrly5Qt++fcnOzrb4PMuWLWPkyJEEBASwceNGi89TWlrKO++8Q1BQEMHBwRb9Pfrhhx8IDQ0FqDbD0qVLGT58OEFBQfz444/mLPeeKuc5fvw4wcHBhIaGMnbsWHJzcwH49NNPCQgIYMSIEXzzzTfmLLdGKme6ZfPmzYwcOdJ025IyVc5z5coVwsPDCQkJISgoiHPnzgEqyKOIv9i2bZui0+kURVGUI0eOKOPHjzdzRQ/mv//9rxIfH68oiqLk5eUpffv2Vd566y1l3759iqIoSkxMjLJ9+3ZzlnjfSkpKlLffflt54YUXlFOnTll0nn379ilvvfWWUl5erhgMBmXx4sUWnUdRFOWrr75SJk+erCiKomRkZCgTJ060yEzLly9Xhg0bpgQGBiqKolSZ4ejRo0poaKhiNBqV8+fPKwEBAeYs+a7uzBMSEqIcO3ZMURRFWbdunZKYmKjk5OQow4YNU4qLi5Vr166Z3larOzMpiqIcO3ZMeeONN0zvs6RMd+bR6XTKF198oSiKonz77bfKN998o4o8cqWhCocOHaJPnz4AdO7cmaNHj5q5ogczePBgpkyZYrpta2tLVlYWzzzzDAD+/v7s3bvXXOU9kAULFhAUFETLli0BLDpPRkYGvr6+TJgwgfHjx9OvXz+LzgPg6elJeXk5RqMRg8GAnZ2dRWZ69NFHWbJkiel2VRkOHTpE79690Wg0tGnThvLycvLy8sxV8l3dmSclJYWOHTsCUF5ejoODAz/++CNdunShUaNGNG7cmEcffZQTJ06Yq+R7ujNTfn4+ycnJREVFmd5nSZnuzHP48GEuXbpEWFgYmzdv5plnnlFFHmkaqmAwGHB1dTXdtrW1payszIwVPRgXFxdcXV0xGAxMnjyZqVOnoigKGo3G9PHCwkIzV1lzer0ed3d3U0MHWHSe/Px8jh49yqJFi5g9ezaRkZEWnQfA2dmZ8+fPM2TIEGJiYggNDbXITIMGDcLOzs50u6oMd/6eUHO2O/PcaroPHz7MJ598QlhYGAaDgcaNG5vu4+LigsFgqPdaa6pypvLycmbOnElUVBQuLi6m+1hSpju/R+fPn8fNzY309HRat27NRx99pIo80jRUwdXVlevXr5tuG43G276ZluTixYu88cYbvPTSS7z44ovY2Pz5Lb9+/Tpubm5mrO7+fPbZZ+zdu5fQ0FCOHz+OTqe77ZGdpeVp2rQpvXv3plGjRnh5eeHg4HDbfzqWlgcgPT2d3r17s23bNj7//HNmzJhBaWmp6eOWmAmo8ufmzt8T169fv+0Xutpt3bqV2NhYli9fjru7u0XnycrK4uzZs8TFxREREcGpU6dISEiw6ExNmzZlwIABAAwYMICjR4+qIo80DVXw8/Nj9+7dAHz//ff4+vqauaIHk5uby5gxY3j33XcZPnw4AE8++STfffcdALt376Zbt27mLPG+rF27lk8++YQ1a9bQsWNHFixYgL+/v8Xm6dq1K3v27EFRFC5dusSNGzd49tlnLTYPgJubm+mXWJMmTSgrK7Pof3O3VJXBz8+PjIwMjEYjFy5cwGg04u7ubuZKa+bzzz83/Sy1a9cOgE6dOnHo0CGKi4spLCwkOzvbYn73derUiS+++II1a9aQkpKCj48PM2fOtOhMXbt2ZdeuXQAcOHAAHx8fVeSxzIfPdez5558nMzOToKAgFEUhMTHR3CU9kA8//JBr166RmppKamoqADNnziQ+Pp6UlBS8vLwYNGiQmat8ODqdjpiYGIvM079/fw4cOMDw4cNRFIVZs2bRtm1bi80DEBYWRlRUFMHBwZSWljJt2jSefvppi84EVf87s7W1pVu3bowcOdK048oSlJeXk5CQQOvWrZk0aRIA3bt3Z/LkyYSGhhIcHIyiKEybNg0HBwczV/twPDw8LDaTTqcjOjqa9evX4+rqynvvvUeTJk3MnkdOuRRCCCFEjcjTE0IIIYSoEWkahBBCCFEj0jQIIYQQokakaRBCCCFEjUjTIIQQQogakaZBiAbi999/x8/Pj9DQUNOfpUuX3vc6GzZsuG1g08OaOHFira1VnQMHDqh2fLAQlkTmNAjRgPj4+LBmzZqHWmPZsmW8/PLLtVQRD9S43K/PPvuMoUOH0qFDhzr/WkJYM2kahBC89957HDhwAEVRCAsLY8iQIezfv9/0H/rNmzdZsGABBw8e5PLly0ybNo1Ro0axfv16Fi5cCECvXr3IzMxkxowZXL16latXr7Js2TJWrFjxl7Uru/V5oaGhPPHEE5w8eRJnZ2e6detGRkYG165dY+XKlezcuZOdO3diMBjIz89nwoQJDBo0iMzMTN5//30cHBxo2rQpiYmJHD9+nOTkZOzt7enZsyd79uwhKysLHx8fvv76a7Zv305ZWRmNGzdmyZIlbNmyhV27dnHz5k3OnTvHuHHjCAgI4IcffiAhIQFFUdBqtSQnJ3P27Fni4+MBTF/PUkYTC/GwpGkQogE5deoUoaGhptvJycmcOHGC33//nfXr11NcXMyIESPo1asXJ0+eJCkpCa1Wy4cffsiXX35JeHg4aWlpLFy4kO+//77ar9OjRw/CwsLYtWtXlWtXd/5Ep06diI6OZuzYsTg6OrJq1Sp0Oh0HDhwAoKioiFWrVpGXl0dgYCADBgwgJiaGdevWodVqWb16NWlpafTr14/i4mI2btwIVDw1M3ToUFq1asXVq1dJT0/HxsaGsWPH8tNPPwEVhxt9/PHHnDlzhvHjxxMQEEBMTAwLFy7E29ubtWvXkp2dzezZs0lMTMTHx4eNGzeyYsUKpk2bVlvfIiFUTZoGIRqQqp6e2LRpE1lZWaZmoqysjAsXLqDVaklISMDZ2ZlLly7h5+d317UrD5f19PQE4Jdffqly7eqahqeeegqoOMPCx8fH9HZxcTFQMe7YxsaGFi1a4ObmRm5uLq6urmi1WtPHU1JS6Nevn6mGymxsbLC3tyciIgJnZ2f++OMP0wm2t566aN26NSUlJQBcuXIFb29vAEJCQgBMjQNAaWlplV9HCGslTYMQDZyXlxf/+Mc/mDt3LkajkdTUVNq2bUtYWBg7duzA1dUVnU5nago0Gg1GoxEHBwcuX74MVBxXC/XtAAABfklEQVTjW1BQYFrz1jHS1a39oLKysoCKw9gMBgMtW7bEYDCQk5NDy5Yt2b9/P+3btwduP5lSo9GgKAonTpxgx44dbNy4kRs3bhAQEHBbrju1bNmSM2fO0L59e5YvX46npyeenp4sWLCANm3acOjQIdPfgRANgTQNQjRwAwYMYP/+/QQHB1NUVMTAgQNxdXXlpZdeYsSIEbi5udGiRQtycnIA6NatG2+++SYrV66kcePGBAYG4u3tXWUzUN3aDyo3N5dRo0ZRWFhIbGwstra2xMfHM2nSJDQaDU2aNGHevHmcPHnyts/7+9//TnJyMikpKTg5OREQEECjRo3w8PAw5arK7NmziYqKwsbGBg8PD8LCwmjdujU6nY7y8nIAEhISHjiPEJZGDqwSQlgEvV7P6dOniYyMNHcpQjRYMqdBCCGEEDUiVxqEEEIIUSNypUEIIYQQNSJNgxBCCCFqRJoGIYQQQtSINA1CCCGEqBFpGoQQQghRI9I0CCGEEKJG/j9tKiIo5ntQKgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 1296x504 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax) = plt.subplots(ncols=1, figsize=(8,5))\n",
    "plt.figure(figsize=(18,7))\n",
    "lgb.plot_importance(lgbm, height=0.8, title=\"Features importance (LightGBM)\", ax=ax,color=\"red\") \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 588,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "lightgbm.basic.Booster"
      ]
     },
     "execution_count": 588,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(lgbm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Random search to fine-tune our hyperparametres"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "https://www.kaggle.com/mlisovyi/lightgbm-hyperparameter-optimisation-lb-0-761"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 807,
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy.stats import randint as sp_randint\n",
    "from scipy.stats import uniform as sp_uniform\n",
    "param_test ={'learning_rate': sp_uniform(loc=0.01, scale=0.19),\n",
    "             'scale_pos_weight': sp_randint(100,150),\n",
    "             'num_leaves': sp_randint(6, 50), \n",
    "             'min_child_samples': sp_randint(100, 500), \n",
    "             #'min_child_weight': [1e-5, 1e-3, 1e-2, 1e-1, 1, 1e1, 1e2, 1e3, 1e4],\n",
    "             'subsample': sp_uniform(loc=0.2, scale=0.8), \n",
    "             'bagging_fraction': sp_uniform(loc=0.4, scale=0.6),\n",
    "             'feature_fraction': sp_uniform(loc=0.4, scale=0.6),\n",
    "             #'reg_alpha': [0, 1e-1, 1, 2, 5, 7, 10, 50, 100],\n",
    "             #'reg_lambda': [0, 1e-1, 1, 5, 10, 20, 50, 100]\n",
    "            }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 826,
   "metadata": {},
   "outputs": [],
   "source": [
    "nb_random = 10\n",
    "\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "clf = lgb.LGBMClassifier(max_depth=-1, \n",
    "                        boosting_type = 'gbdt',\n",
    "                        objective= 'binary',\n",
    "                        random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 845,
   "metadata": {},
   "outputs": [],
   "source": [
    "random_search = RandomizedSearchCV(estimator=clf,\n",
    "                                  param_distributions=param_test,\n",
    "                                  n_iter=nb_random,\n",
    "                                  scoring='balanced_accuracy',\n",
    "                                  cv=3,\n",
    "                                  refit=True,\n",
    "                                  random_state=42,\n",
    "                                  verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 846,
   "metadata": {},
   "outputs": [],
   "source": [
    "categorical_features = ['gender', 'education', 'marriage', 'pay_1', 'pay_2', 'pay_3', 'pay_4', 'pay_5', 'pay_6']\n",
    "\n",
    "fit_params={\"early_stopping_rounds\": 30, \n",
    "            \"eval_metric\" : 'balanced_accuracy', \n",
    "            \"eval_set\" : [(X_raw_test, np.ravel(y_test))],\n",
    "            'eval_names': ['valid'],\n",
    "            'verbose': 100,\n",
    "            'categorical_feature': categorical_features}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 847,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 848,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 10 candidates, totalling 30 fits\n",
      "Training until validation scores don't improve for 30 rounds\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\lightgbm\\basic.py:1247: UserWarning: categorical_feature in Dataset is overridden.\n",
      "New categorical_feature is ['education', 'gender', 'marriage', 'pay_1', 'pay_2', 'pay_3', 'pay_4', 'pay_5', 'pay_6']\n",
      "  'New categorical_feature is {}'.format(sorted(list(categorical_feature))))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.5593\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.559155\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.559181\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.530263\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.530482\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.530023\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.526354\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.526353\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.526366\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.528228\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.528268\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.528362\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.550725\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.550892\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.550656\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587646\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587387\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587645\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.527089\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.527143\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.526933\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.532241\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.533124\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.532882\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.584626\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.584858\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.584017\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587625\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587548\n",
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.587106\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  30 out of  30 | elapsed:    9.7s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 30 rounds\n",
      "Early stopping, best iteration is:\n",
      "[1]\tvalid's binary_logloss: 0.559521\n",
      "Random search with 10 tries took: 10.39s\n"
     ]
    }
   ],
   "source": [
    "t0 = time.time()\n",
    "random_search.fit(X_raw, np.ravel(y_raw), **fit_params)\n",
    "t1 = time.time()\n",
    "print('Random search with {} tries took: {:.2f}s'.format(nb_random, t1-t0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 849,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt_params = random_search.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 850,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5"
      ]
     },
     "execution_count": 850,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_search.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 851,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_random_search = random_search.predict(X_raw_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 852,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_random_search = np.where(y_pred_random_search>0.5, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 853,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.78      1.00      0.88      5873\n",
      "           1       0.00      0.00      0.00      1627\n",
      "\n",
      "    accuracy                           0.78      7500\n",
      "   macro avg       0.39      0.50      0.44      7500\n",
      "weighted avg       0.61      0.78      0.69      7500\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\sklearn\\metrics\\classification.py:1437: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples.\n",
      "  'precision', 'predicted', average, warn_for)\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, y_pred_random_search))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 605,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = lgb.LGBMClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 254,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "metadata": {},
   "outputs": [],
   "source": [
    "categorical_features = ['gender', 'education', 'marriage', 'pay_1', 'pay_2', 'pay_3', 'pay_4', 'pay_5', 'pay_6']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 307,
   "metadata": {},
   "outputs": [],
   "source": [
    "params_lgbm = {\n",
    "          'boosting_type': 'gbdt',\n",
    "          'objective': 'binary',\n",
    "            'is_unbalanced':True,\n",
    "          'metric':'auc',\n",
    "          'learning_rate': 0.05,\n",
    "          'num_leaves': 15,  # we should let it be smaller than 2^(max_depth)\n",
    "          'max_depth': 7,  # -1 means no limit\n",
    "          'min_child_samples': 100,  # Minimum number of data need in a child(min_data_in_leaf)\n",
    "          'max_bin': 100,  # Number of bucketed bin for feature values\n",
    "          'subsample': 0.9,  # Subsample ratio of the training instance.\n",
    "          'subsample_freq': 1,  # frequence of subsample, <=0 means no enable\n",
    "          'colsample_bytree': 0.7,  # Subsample ratio of columns when constructing each tree.\n",
    "          'min_child_weight': 0,  # Minimum sum of instance weight(hessian) needed in a child(leaf)\n",
    "          'min_split_gain': 0,  # lambda_l1, lambda_l2 and min_gain_to_split to regularization\n",
    "          'nthread': 8,\n",
    "          'verbose': 1,\n",
    "          'scale_pos_weight':50, # because training data is sightly unbalanced \n",
    "         }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 308,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 1, 0, ..., 1, 1, 1], dtype=int64)"
      ]
     },
     "execution_count": 308,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.ravel(y_raw.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 309,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "limit_bal    float64\n",
       "gender         int64\n",
       "education      int64\n",
       "marriage       int64\n",
       "age            int64\n",
       "pay_1          int64\n",
       "pay_2          int64\n",
       "pay_3          int64\n",
       "pay_4          int64\n",
       "pay_5          int64\n",
       "pay_6          int64\n",
       "bill_amt1    float64\n",
       "bill_amt2    float64\n",
       "bill_amt3    float64\n",
       "bill_amt4    float64\n",
       "bill_amt5    float64\n",
       "bill_amt6    float64\n",
       "pay_amt1     float64\n",
       "pay_amt2     float64\n",
       "pay_amt3     float64\n",
       "pay_amt4     float64\n",
       "pay_amt5     float64\n",
       "pay_amt6     float64\n",
       "dtype: object"
      ]
     },
     "execution_count": 309,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# on vérifie que les types des variable catégorielles sont bien des int, et non pas categorical\n",
    "X_raw.dtypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 310,
   "metadata": {},
   "outputs": [],
   "source": [
    "dtrain = lgb.Dataset(X_raw.values, \n",
    "                     label=np.ravel(y_raw.values),\n",
    "                     feature_name=list(X_raw.columns),\n",
    "                     categorical_feature=categorical_features,\n",
    "                    free_raw_data=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 311,
   "metadata": {},
   "outputs": [],
   "source": [
    "evals_results={}\n",
    "MAX_ROUNDS = 1000 #lgb iterations\n",
    "EARLY_STOP = 50 #lgb early stop \n",
    "OPT_ROUNDS = 1000  #To be adjusted based on best validation rounds\n",
    "VERBOSE_EVAL = 50 #Print out metric result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 312,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\lightgbm\\basic.py:1243: UserWarning: Using categorical_feature in Dataset.\n",
      "  warnings.warn('Using categorical_feature in Dataset.')\n"
     ]
    }
   ],
   "source": [
    "lgbm = lgb.train(params_lgbm, \n",
    "                  dtrain, \n",
    "                  evals_result=evals_results, \n",
    "                  num_boost_round=MAX_ROUNDS,\n",
    "                  verbose_eval=VERBOSE_EVAL, \n",
    "                  feval=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 313,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>default</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>30000.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.221200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.415062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            default\n",
       "count  30000.000000\n",
       "mean       0.221200\n",
       "std        0.415062\n",
       "min        0.000000\n",
       "25%        0.000000\n",
       "50%        0.000000\n",
       "75%        0.000000\n",
       "max        1.000000"
      ]
     },
     "execution_count": 313,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_raw.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 322,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = lgbm.predict(X_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 323,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.98565234, 0.97167218, 0.82523721, ..., 0.95817876, 0.93619706,\n",
       "       0.87557382])"
      ]
     },
     "execution_count": 323,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 324,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = np.where(y_pred>0.5, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 328,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 5215, 18149],\n",
       "       [    0,  6636]], dtype=int64)"
      ]
     },
     "execution_count": 328,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_raw, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 326,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.22      0.36     23364\n",
      "           1       0.27      1.00      0.42      6636\n",
      "\n",
      "    accuracy                           0.40     30000\n",
      "   macro avg       0.63      0.61      0.39     30000\n",
      "weighted avg       0.84      0.40      0.38     30000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_raw, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 327,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, (ax) = plt.subplots(ncols=1, figsize=(8,5))\n",
    "lgb.plot_importance(lgbm, height=0.8, title=\"Features importance (LightGBM)\", ax=ax,color=\"red\") \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 287,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.98617058, 0.97296606, 0.86569794, ..., 0.98072076, 0.90652921,\n",
       "       0.90622171])"
      ]
     },
     "execution_count": 287,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lgbm.predict(X_raw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "categorical_features = ['SEX', 'EDUCATION', 'MARRIAGE','PAY_0', 'PAY_2', 'PAY_3', 'PAY_4', 'PAY_5', 'PAY_6']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CatBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pipeline Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import FunctionTransformer\n",
    "bin_age_transformer = FunctionTransformer(bin_age, validate=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, y = load_data_xy(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>limit_bal</th>\n",
       "      <th>gender</th>\n",
       "      <th>education</th>\n",
       "      <th>marriage</th>\n",
       "      <th>age</th>\n",
       "      <th>pay_1</th>\n",
       "      <th>pay_2</th>\n",
       "      <th>pay_3</th>\n",
       "      <th>pay_4</th>\n",
       "      <th>pay_5</th>\n",
       "      <th>pay_6</th>\n",
       "      <th>bill_amt1</th>\n",
       "      <th>bill_amt2</th>\n",
       "      <th>bill_amt3</th>\n",
       "      <th>bill_amt4</th>\n",
       "      <th>bill_amt5</th>\n",
       "      <th>bill_amt6</th>\n",
       "      <th>pay_amt1</th>\n",
       "      <th>pay_amt2</th>\n",
       "      <th>pay_amt3</th>\n",
       "      <th>pay_amt4</th>\n",
       "      <th>pay_amt5</th>\n",
       "      <th>pay_amt6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>24</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3913.0</td>\n",
       "      <td>3102.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>120000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>1725.0</td>\n",
       "      <td>2682.0</td>\n",
       "      <td>3272.0</td>\n",
       "      <td>3455.0</td>\n",
       "      <td>3261.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>90000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>29239.0</td>\n",
       "      <td>14027.0</td>\n",
       "      <td>13559.0</td>\n",
       "      <td>14331.0</td>\n",
       "      <td>14948.0</td>\n",
       "      <td>15549.0</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>1500.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>5000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>46990.0</td>\n",
       "      <td>48233.0</td>\n",
       "      <td>49291.0</td>\n",
       "      <td>28314.0</td>\n",
       "      <td>28959.0</td>\n",
       "      <td>29547.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>1100.0</td>\n",
       "      <td>1069.0</td>\n",
       "      <td>1000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>50000.0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>8617.0</td>\n",
       "      <td>5670.0</td>\n",
       "      <td>35835.0</td>\n",
       "      <td>20940.0</td>\n",
       "      <td>19146.0</td>\n",
       "      <td>19131.0</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>36681.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>9000.0</td>\n",
       "      <td>689.0</td>\n",
       "      <td>679.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   limit_bal  gender  education  marriage  age  pay_1  pay_2  pay_3  pay_4  \\\n",
       "0    20000.0       2          2         1   24      1      1      0      0   \n",
       "1   120000.0       2          2         2   26      0      1      0      0   \n",
       "2    90000.0       2          2         2   34      0      0      0      0   \n",
       "3    50000.0       2          2         1   37      0      0      0      0   \n",
       "4    50000.0       1          2         1   57      0      0      0      0   \n",
       "\n",
       "   pay_5  pay_6  bill_amt1  bill_amt2  bill_amt3  bill_amt4  bill_amt5  \\\n",
       "0      0      0     3913.0     3102.0      689.0        0.0        0.0   \n",
       "1      0      1     2682.0     1725.0     2682.0     3272.0     3455.0   \n",
       "2      0      0    29239.0    14027.0    13559.0    14331.0    14948.0   \n",
       "3      0      0    46990.0    48233.0    49291.0    28314.0    28959.0   \n",
       "4      0      0     8617.0     5670.0    35835.0    20940.0    19146.0   \n",
       "\n",
       "   bill_amt6  pay_amt1  pay_amt2  pay_amt3  pay_amt4  pay_amt5  pay_amt6  \n",
       "0        0.0       0.0     689.0       0.0       0.0       0.0       0.0  \n",
       "1     3261.0       0.0    1000.0    1000.0    1000.0       0.0    2000.0  \n",
       "2    15549.0    1518.0    1500.0    1000.0    1000.0    1000.0    5000.0  \n",
       "3    29547.0    2000.0    2019.0    1200.0    1100.0    1069.0    1000.0  \n",
       "4    19131.0    2000.0   36681.0   10000.0    9000.0     689.0     679.0  "
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'feature_eng'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-88-b626cf4277a1>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0mdataprep\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpipelines\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mAgeBinAdder\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0mAgeBinAdder\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtransform\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mX\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\UCI-Credit-card-defaults\\dataprep\\pipelines.py\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbase\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mBaseEstimator\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mTransformerMixin\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpreprocessing\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mFunctionTransformer\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 6\u001b[1;33m \u001b[1;32mimport\u001b[0m \u001b[0mfeature_eng\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      7\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      8\u001b[0m \u001b[1;31m# Age Bin Adder\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mModuleNotFoundError\u001b[0m: No module named 'feature_eng'"
     ]
    }
   ],
   "source": [
    "from dataprep.pipelines import AgeBinAdder\n",
    "AgeBinAdder().transform(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pourquoi le module n'est pas trouvable?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TO DO:\n",
    "create a pipeline for different processings of the dataset: 1 is categorical, while the other one is 1-hot encoded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Relève la tête du guidon gros!\n",
    "## Quel est le Big Picture du problème?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Notre objectif = prédire quelles sont les personnes les plus susceptibles de défault  \n",
    "ie trouver celles pour qui y=1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Quelles métriques sont possibles?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Classification accuracy: no beacuase of our imbalanced classes to predict\n",
    "* Logarithmic loss: \n",
    "* Area Under ROC Curve: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Quelles métriques pour cet objectif?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* recall = parmi les (y=1), quelle proportion a bien été détectée?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C:\\Users\\twang\\AppData\\Local\\Continuum\\anaconda3\\python.exe\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "print(sys.executable)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xgboost as xgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import lightgbm as lgb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  },
  "toc-autonumbering": true
 },
 "nbformat": 4,
 "nbformat_minor": 4
}