lgbm.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Import Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns \n",
    "from matplotlib import pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import roc_auc_score\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "pd.set_option('display.max_columns', 100)\n",
    "from sklearn.preprocessing import *\n",
    "from sklearn.model_selection import *\n",
    "from sklearn.metrics import *\n",
    "import gc\n",
    "from tqdm import tqdm, tqdm_notebook"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('Data/Train/Train.csv')\n",
    "sub = pd.read_csv('Data/sample_submmission.csv')\n",
    "test = pd.read_csv('Data/test_l0Auv8Q.csv')\n",
    "fh = pd.read_csv('Data/Train/First_Health_Camp_Attended.csv')\n",
    "sh = pd.read_csv('Data/Train/Second_Health_Camp_Attended.csv')\n",
    "th = pd.read_csv('Data/Train/Third_Health_Camp_Attended.csv')\n",
    "pp = pd.read_csv('Data/Train/Patient_Profile.csv')\n",
    "hc = pd.read_csv('Data/Train/Health_Camp_Detail.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Registration_Date</th>\n",
       "      <th>Var1</th>\n",
       "      <th>Var2</th>\n",
       "      <th>Var3</th>\n",
       "      <th>Var4</th>\n",
       "      <th>Var5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>489652</td>\n",
       "      <td>6578</td>\n",
       "      <td>10-Sep-05</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>507246</td>\n",
       "      <td>6578</td>\n",
       "      <td>18-Aug-05</td>\n",
       "      <td>45</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID Registration_Date  Var1  Var2  Var3  Var4  Var5\n",
       "0      489652            6578         10-Sep-05     4     0     0     0     2\n",
       "1      507246            6578         18-Aug-05    45     5     0     0     7"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Registration_Date</th>\n",
       "      <th>Var1</th>\n",
       "      <th>Var2</th>\n",
       "      <th>Var3</th>\n",
       "      <th>Var4</th>\n",
       "      <th>Var5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>505701</td>\n",
       "      <td>6548</td>\n",
       "      <td>21-May-06</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>500633</td>\n",
       "      <td>6584</td>\n",
       "      <td>02-Jun-06</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": [
       "   Patient_ID  Health_Camp_ID Registration_Date  Var1  Var2  Var3  Var4  Var5\n",
       "0      505701            6548         21-May-06     1     0     0     0     2\n",
       "1      500633            6584         02-Jun-06     0     0     0     0     0"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Outcome</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>505701</td>\n",
       "      <td>6548</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>500633</td>\n",
       "      <td>6584</td>\n",
       "      <td>0.5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID  Outcome\n",
       "0      505701            6548      0.5\n",
       "1      500633            6584      0.5"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Donation</th>\n",
       "      <th>Health_Score</th>\n",
       "      <th>Unnamed: 4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>506181</td>\n",
       "      <td>6560</td>\n",
       "      <td>40</td>\n",
       "      <td>0.439024</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>494977</td>\n",
       "      <td>6560</td>\n",
       "      <td>20</td>\n",
       "      <td>0.097561</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID  Donation  Health_Score  Unnamed: 4\n",
       "0      506181            6560        40      0.439024         NaN\n",
       "1      494977            6560        20      0.097561         NaN"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fh.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Health Score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>526631</td>\n",
       "      <td>6536</td>\n",
       "      <td>0.875136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>509122</td>\n",
       "      <td>6536</td>\n",
       "      <td>0.755700</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID  Health Score\n",
       "0      526631            6536      0.875136\n",
       "1      509122            6536      0.755700"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sh.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Number_of_stall_visited</th>\n",
       "      <th>Last_Stall_Visited_Number</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>517875</td>\n",
       "      <td>6527</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>504692</td>\n",
       "      <td>6578</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID  Number_of_stall_visited  \\\n",
       "0      517875            6527                        3   \n",
       "1      504692            6578                        1   \n",
       "\n",
       "   Last_Stall_Visited_Number  \n",
       "0                          1  \n",
       "1                          1  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "th.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>Health_Camp_ID</th>\n",
       "      <th>Camp_Start_Date</th>\n",
       "      <th>Camp_End_Date</th>\n",
       "      <th>Category1</th>\n",
       "      <th>Category2</th>\n",
       "      <th>Category3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6560</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>20-Aug-03</td>\n",
       "      <td>First</td>\n",
       "      <td>B</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6530</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>28-Oct-03</td>\n",
       "      <td>First</td>\n",
       "      <td>C</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Health_Camp_ID Camp_Start_Date Camp_End_Date Category1 Category2  Category3\n",
       "0            6560       16-Aug-03     20-Aug-03     First         B          2\n",
       "1            6530       16-Aug-03     28-Oct-03     First         C          2"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hc.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>Patient_ID</th>\n",
       "      <th>Online_Follower</th>\n",
       "      <th>LinkedIn_Shared</th>\n",
       "      <th>Twitter_Shared</th>\n",
       "      <th>Facebook_Shared</th>\n",
       "      <th>Income</th>\n",
       "      <th>Education_Score</th>\n",
       "      <th>Age</th>\n",
       "      <th>First_Interaction</th>\n",
       "      <th>City_Type</th>\n",
       "      <th>Employer_Category</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>516956</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>39</td>\n",
       "      <td>18-Jun-03</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Software Industry</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>507733</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>40</td>\n",
       "      <td>20-Jul-03</td>\n",
       "      <td>H</td>\n",
       "      <td>Software Industry</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Online_Follower  LinkedIn_Shared  Twitter_Shared  \\\n",
       "0      516956                0                0               0   \n",
       "1      507733                0                0               0   \n",
       "\n",
       "   Facebook_Shared Income Education_Score Age First_Interaction City_Type  \\\n",
       "0                0      1              90  39         18-Jun-03       NaN   \n",
       "1                0      1            None  40         20-Jul-03         H   \n",
       "\n",
       "   Employer_Category  \n",
       "0  Software Industry  \n",
       "1  Software Industry  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pp.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "pp['online_Activity_Score']=pp['Online_Follower']+pp['LinkedIn_Shared']+pp['Twitter_Shared']+pp['Facebook_Shared']\n",
    "del pp['Online_Follower']\n",
    "del pp['LinkedIn_Shared']\n",
    "del pp['Twitter_Shared']\n",
    "del pp['Facebook_Shared']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((75278, 8),\n",
       " (35249, 8),\n",
       " (35249, 3),\n",
       " (6218, 5),\n",
       " (7819, 3),\n",
       " (6515, 4),\n",
       " (37633, 8),\n",
       " (65, 6))"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.shape,test.shape,sub.shape,fh.shape,sh.shape,th.shape,pp.shape,hc.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing and feature engineering patient profile"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "pp[['Income', 'Education_Score', 'Age']] = pp[['Income', 'Education_Score', 'Age']].apply(lambda x: x.str.replace('None', 'NaN').astype('float'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "le = LabelEncoder()\n",
    "\n",
    "for col in ['City_Type','Employer_Category']:\n",
    "    pp[col]=  pp[col].astype('str')\n",
    "    pp[col]= le.fit_transform(pp[col]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "pp['first_interaction_year'] = pd.to_datetime(pp['First_Interaction']).dt.year\n",
    "pp['first_interaction_month'] = pd.to_datetime(pp['First_Interaction']).dt.month\n",
    "pp['first_interaction_day'] = pd.to_datetime(pp['First_Interaction']).dt.day"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>Patient_ID</th>\n",
       "      <th>Income</th>\n",
       "      <th>Education_Score</th>\n",
       "      <th>Age</th>\n",
       "      <th>First_Interaction</th>\n",
       "      <th>City_Type</th>\n",
       "      <th>Employer_Category</th>\n",
       "      <th>online_Activity_Score</th>\n",
       "      <th>first_interaction_year</th>\n",
       "      <th>first_interaction_month</th>\n",
       "      <th>first_interaction_day</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>516956</td>\n",
       "      <td>1.0</td>\n",
       "      <td>90.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>18-Jun-03</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>2003</td>\n",
       "      <td>6</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>507733</td>\n",
       "      <td>1.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>40.0</td>\n",
       "      <td>20-Jul-03</td>\n",
       "      <td>7</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>2003</td>\n",
       "      <td>7</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Income  Education_Score   Age First_Interaction  City_Type  \\\n",
       "0      516956     1.0             90.0  39.0         18-Jun-03          9   \n",
       "1      507733     1.0              NaN  40.0         20-Jul-03          7   \n",
       "\n",
       "   Employer_Category  online_Activity_Score  first_interaction_year  \\\n",
       "0                 10                      0                    2003   \n",
       "1                 10                      0                    2003   \n",
       "\n",
       "   first_interaction_month  first_interaction_day  \n",
       "0                        6                     18  \n",
       "1                        7                     20  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pp.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing and feature engineering Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['registration_year'] = pd.to_datetime(train['Registration_Date']).dt.year\n",
    "train['registration_month'] = pd.to_datetime(train['Registration_Date']).dt.month\n",
    "train['registration_day'] = pd.to_datetime(train['Registration_Date']).dt.day"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['Registration_Date']=pd.to_datetime(train['Registration_Date'])\n",
    "train.sort_values(by=['Patient_ID','Registration_Date'],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['days_since_last_registration'] = train.groupby('Patient_ID')['Registration_Date'].diff().apply(lambda x: x.days)\n",
    "train['days_since_next_registration'] = train.groupby('Patient_ID')['Registration_Date'].diff(-1) * (-1) / np.timedelta64(1, 'D')       \n",
    "train.reset_index(drop=True,inplace=True)                                                                                                      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['Unique_Health_Camp_per_patient']=train.groupby(['Patient_ID'])['Health_Camp_ID'].transform('nunique')\n",
    "train['Unique_Patient_per_HealthCamp']=train.groupby(['Health_Camp_ID'])['Patient_ID'].transform('nunique')\n",
    "train['Unique_year_per_patient']=train.groupby(['Patient_ID'])['registration_year'].transform('nunique')\n",
    "train['Unique_months_per_patient']=train.groupby(['Patient_ID'])['registration_month'].transform('nunique')\n",
    "train['Unique_day_per_patient']=train.groupby(['Patient_ID'])['registration_day'].transform('nunique')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Registration_Date</th>\n",
       "      <th>Var1</th>\n",
       "      <th>Var2</th>\n",
       "      <th>Var3</th>\n",
       "      <th>Var4</th>\n",
       "      <th>Var5</th>\n",
       "      <th>registration_year</th>\n",
       "      <th>registration_month</th>\n",
       "      <th>registration_day</th>\n",
       "      <th>days_since_last_registration</th>\n",
       "      <th>days_since_next_registration</th>\n",
       "      <th>Unique_Health_Camp_per_patient</th>\n",
       "      <th>Unique_Patient_per_HealthCamp</th>\n",
       "      <th>Unique_year_per_patient</th>\n",
       "      <th>Unique_months_per_patient</th>\n",
       "      <th>Unique_day_per_patient</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>485679</td>\n",
       "      <td>6578</td>\n",
       "      <td>2005-08-22</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2005.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>9.0</td>\n",
       "      <td>2</td>\n",
       "      <td>2837</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>485679</td>\n",
       "      <td>6555</td>\n",
       "      <td>2005-08-31</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2005.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2</td>\n",
       "      <td>1738</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID Registration_Date  Var1  Var2  Var3  Var4  Var5  \\\n",
       "0      485679            6578        2005-08-22     0     0     0     0     0   \n",
       "1      485679            6555        2005-08-31     0     0     0     0     0   \n",
       "\n",
       "   registration_year  registration_month  registration_day  \\\n",
       "0             2005.0                 8.0              22.0   \n",
       "1             2005.0                 8.0              31.0   \n",
       "\n",
       "   days_since_last_registration  days_since_next_registration  \\\n",
       "0                           NaN                           9.0   \n",
       "1                           9.0                           NaN   \n",
       "\n",
       "   Unique_Health_Camp_per_patient  Unique_Patient_per_HealthCamp  \\\n",
       "0                               2                           2837   \n",
       "1                               2                           1738   \n",
       "\n",
       "   Unique_year_per_patient  Unique_months_per_patient  Unique_day_per_patient  \n",
       "0                      1.0                        1.0                     2.0  \n",
       "1                      1.0                        1.0                     2.0  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing and feature engineering Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['registration_year'] = pd.to_datetime(test['Registration_Date']).dt.year\n",
    "test['registration_month'] = pd.to_datetime(test['Registration_Date']).dt.month\n",
    "test['registration_day'] = pd.to_datetime(test['Registration_Date']).dt.day"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['Registration_Date']=pd.to_datetime(test['Registration_Date'])\n",
    "test.sort_values(by=['Patient_ID','Registration_Date'],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['days_since_last_registration'] = test.groupby('Patient_ID')['Registration_Date'].diff().apply(lambda x: x.days)\n",
    "test['days_since_next_registration'] = test.groupby('Patient_ID')['Registration_Date'].diff(-1) * (-1) / np.timedelta64(1, 'D')\n",
    "test.reset_index(drop=True,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['Unique_Health_Camp_per_patient']=test.groupby(['Patient_ID'])['Health_Camp_ID'].transform('nunique')\n",
    "test['Unique_Patient_per_HealthCamp']=test.groupby(['Health_Camp_ID'])['Patient_ID'].transform('nunique')\n",
    "test['Unique_year_per_patient']=test.groupby(['Patient_ID'])['registration_year'].transform('nunique')\n",
    "test['Unique_months_per_patient']=test.groupby(['Patient_ID'])['registration_month'].transform('nunique')\n",
    "test['Unique_day_per_patient']=test.groupby(['Patient_ID'])['registration_day'].transform('nunique')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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>Patient_ID</th>\n",
       "      <th>Health_Camp_ID</th>\n",
       "      <th>Registration_Date</th>\n",
       "      <th>Var1</th>\n",
       "      <th>Var2</th>\n",
       "      <th>Var3</th>\n",
       "      <th>Var4</th>\n",
       "      <th>Var5</th>\n",
       "      <th>registration_year</th>\n",
       "      <th>registration_month</th>\n",
       "      <th>registration_day</th>\n",
       "      <th>days_since_last_registration</th>\n",
       "      <th>days_since_next_registration</th>\n",
       "      <th>Unique_Health_Camp_per_patient</th>\n",
       "      <th>Unique_Patient_per_HealthCamp</th>\n",
       "      <th>Unique_year_per_patient</th>\n",
       "      <th>Unique_months_per_patient</th>\n",
       "      <th>Unique_day_per_patient</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>485678</td>\n",
       "      <td>6551</td>\n",
       "      <td>2006-11-13</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2006</td>\n",
       "      <td>11</td>\n",
       "      <td>13</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>2180</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>485683</td>\n",
       "      <td>6584</td>\n",
       "      <td>2006-05-29</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2006</td>\n",
       "      <td>5</td>\n",
       "      <td>29</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5</td>\n",
       "      <td>3041</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Patient_ID  Health_Camp_ID Registration_Date  Var1  Var2  Var3  Var4  Var5  \\\n",
       "0      485678            6551        2006-11-13     0     0     0     0     0   \n",
       "1      485683            6584        2006-05-29     0     0     0     0     0   \n",
       "\n",
       "   registration_year  registration_month  registration_day  \\\n",
       "0               2006                  11                13   \n",
       "1               2006                   5                29   \n",
       "\n",
       "   days_since_last_registration  days_since_next_registration  \\\n",
       "0                           NaN                           NaN   \n",
       "1                           NaN                           0.0   \n",
       "\n",
       "   Unique_Health_Camp_per_patient  Unique_Patient_per_HealthCamp  \\\n",
       "0                               1                           2180   \n",
       "1                               5                           3041   \n",
       "\n",
       "   Unique_year_per_patient  Unique_months_per_patient  Unique_day_per_patient  \n",
       "0                        1                          1                       1  \n",
       "1                        2                          2                       2  "
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### find overlap between train and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total uniquePatient_ID values in Train are 29828\n",
      "Total uniquePatient_ID values in Test are 15324\n",
      "CommonPatient_ID values are 8460\n",
      "**************************\n"
     ]
    }
   ],
   "source": [
    "cols =  ['Patient_ID']\n",
    "for col in cols:\n",
    "  print('Total unique'+col  +' values in Train are {}'.format(train[col].nunique()))\n",
    "  print('Total unique'+col  +' values in Test are {}'.format(test[col].nunique()))\n",
    "  print('Common'+col +' values are {}'.format(len(list(set(train[col]) & set(test[col])))))\n",
    "  print('**************************')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Merging train,test with Patient Profile"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.merge(train, pp, on = 'Patient_ID', how = 'left')\n",
    "test = pd.merge(test, pp, on = 'Patient_ID', how = 'left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Getting Target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "for c in [fh, sh, th, train]:\n",
    "  c['id'] = c['Patient_ID'].astype('str') + c['Health_Camp_ID'].astype('str')\n",
    "th = th[th['Number_of_stall_visited'] > 0]\n",
    "\n",
    "all_patients_in_camp = pd.Series(fh['id'].tolist() + sh['id'].tolist() + th['id'].tolist()).unique()\n",
    "\n",
    "train['target'] = 0\n",
    "train.loc[train['id'].isin(all_patients_in_camp), 'target'] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    54744\n",
       "1    20534\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train['target'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing and feature engineering first health camp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# groupeddf=fh.groupby('Patient_ID', as_index=False)['Donation'].agg('sum').reset_index(drop=True)\n",
    "# groupeddf=groupeddf.rename(columns={'Donation':'Donation_Sum'})\n",
    "# groupeddf=groupeddf[groupeddf['Patient_ID'].isin(test['Patient_ID'])]\n",
    "# train=pd.merge(train,groupeddf,on=['Patient_ID'],how='left')\n",
    "# train['Donation_Sum'].fillna(-99,inplace=True)\n",
    "# test=pd.merge(test,groupeddf,on=['Patient_ID'],how='left')\n",
    "# test['Donation_Sum'].fillna(-99,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### preprocessing and feature engineering third health camp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# groupeddf=th.groupby('Patient_ID', as_index=False)['Number_of_stall_visited'].agg('sum').reset_index(drop=True)\n",
    "# groupeddf=groupeddf.rename(columns={'Number_of_stall_visited':'Total_Stalls_Visited'})\n",
    "# groupeddf=groupeddf[groupeddf['Patient_ID'].isin(test['Patient_ID'])]\n",
    "# train=pd.merge(train,groupeddf,on=['Patient_ID'],how='left')\n",
    "# train['Total_Stalls_Visited'].fillna(-99,inplace=True)\n",
    "# test=pd.merge(test,groupeddf,on=['Patient_ID'],how='left')\n",
    "# test['Total_Stalls_Visited'].fillna(-99,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Preprocessing and feature engineering health camp details"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>Health_Camp_ID</th>\n",
       "      <th>Camp_Start_Date</th>\n",
       "      <th>Camp_End_Date</th>\n",
       "      <th>Category1</th>\n",
       "      <th>Category2</th>\n",
       "      <th>Category3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6560</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>20-Aug-03</td>\n",
       "      <td>First</td>\n",
       "      <td>B</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6530</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>28-Oct-03</td>\n",
       "      <td>First</td>\n",
       "      <td>C</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Health_Camp_ID Camp_Start_Date Camp_End_Date Category1 Category2  Category3\n",
       "0            6560       16-Aug-03     20-Aug-03     First         B          2\n",
       "1            6530       16-Aug-03     28-Oct-03     First         C          2"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hc.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def timediff(duration):\n",
    "    duration_in_s = duration.total_seconds()\n",
    "    days = divmod(duration_in_s, 86400)[0]\n",
    "    return days"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "hc['Camp_Duration']=pd.to_datetime(hc['Camp_End_Date'])-pd.to_datetime(hc['Camp_Start_Date'])\n",
    "hc['Camp_Duration']=hc['Camp_Duration'].apply(timediff)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "hc['camp_start_year'] = pd.to_datetime(hc['Camp_Start_Date']).dt.year\n",
    "hc['camp_start_month'] = pd.to_datetime(hc['Camp_Start_Date']).dt.month\n",
    "hc['camp_start_day'] = pd.to_datetime(hc['Camp_Start_Date']).dt.day\n",
    "\n",
    "hc['camp_end_year'] = pd.to_datetime(hc['Camp_End_Date']).dt.year\n",
    "hc['camp_end_month'] = pd.to_datetime(hc['Camp_End_Date']).dt.month\n",
    "hc['camp_end_day'] = pd.to_datetime(hc['Camp_End_Date']).dt.day"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "hc['Category1'] = hc['Category1'].map({'First': 1, 'Second': 2, 'Third': 3})\n",
    "hc['Category2'] = pd.factorize(hc['Category2'])[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>Health_Camp_ID</th>\n",
       "      <th>Camp_Start_Date</th>\n",
       "      <th>Camp_End_Date</th>\n",
       "      <th>Category1</th>\n",
       "      <th>Category2</th>\n",
       "      <th>Category3</th>\n",
       "      <th>Camp_Duration</th>\n",
       "      <th>camp_start_year</th>\n",
       "      <th>camp_start_month</th>\n",
       "      <th>camp_start_day</th>\n",
       "      <th>camp_end_year</th>\n",
       "      <th>camp_end_month</th>\n",
       "      <th>camp_end_day</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6560</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>20-Aug-03</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>4.0</td>\n",
       "      <td>2003</td>\n",
       "      <td>8</td>\n",
       "      <td>16</td>\n",
       "      <td>2003</td>\n",
       "      <td>8</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6530</td>\n",
       "      <td>16-Aug-03</td>\n",
       "      <td>28-Oct-03</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>73.0</td>\n",
       "      <td>2003</td>\n",
       "      <td>8</td>\n",
       "      <td>16</td>\n",
       "      <td>2003</td>\n",
       "      <td>10</td>\n",
       "      <td>28</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Health_Camp_ID Camp_Start_Date Camp_End_Date  Category1  Category2  \\\n",
       "0            6560       16-Aug-03     20-Aug-03          1          0   \n",
       "1            6530       16-Aug-03     28-Oct-03          1          1   \n",
       "\n",
       "   Category3  Camp_Duration  camp_start_year  camp_start_month  \\\n",
       "0          2            4.0             2003                 8   \n",
       "1          2           73.0             2003                 8   \n",
       "\n",
       "   camp_start_day  camp_end_year  camp_end_month  camp_end_day  \n",
       "0              16           2003               8            20  \n",
       "1              16           2003              10            28  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hc.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Merging train,test with health camp details"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.merge(train, hc, on = 'Health_Camp_ID', how = 'left')\n",
    "test = pd.merge(test, hc, on = 'Health_Camp_ID', how = 'left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### More Feature Engineering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['Unique_camp_year_per_patient']=train.groupby(['Patient_ID'])['camp_start_year'].transform('nunique')\n",
    "train['Unique_camp_months_per_patient']=train.groupby(['Patient_ID'])['camp_start_month'].transform('nunique')\n",
    "train['Unique_camp_day_per_patient']=train.groupby(['Patient_ID'])['camp_start_day'].transform('nunique')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "test['Unique_camp_year_per_patient']=test.groupby(['Patient_ID'])['camp_start_year'].transform('nunique')\n",
    "test['Unique_camp_months_per_patient']=test.groupby(['Patient_ID'])['camp_start_month'].transform('nunique')\n",
    "test['Unique_camp_day_per_patient']=test.groupby(['Patient_ID'])['camp_start_day'].transform('nunique')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# groupeddf=train.groupby('Patient_ID', as_index=False)['target'].agg('mean').reset_index(drop=True)\n",
    "# groupeddf=groupeddf[groupeddf['Patient_ID'].isin(test['Patient_ID'])]\n",
    "# groupeddf.rename(columns={'target':'meantarget'},inplace=True)\n",
    "# train=pd.merge(train,groupeddf,on=['Patient_ID'],how='left')\n",
    "# train['meantarget'].fillna(-99,inplace=True)\n",
    "# test=pd.merge(test,groupeddf,on=['Patient_ID'],how='left')\n",
    "# test['meantarget'].fillna(-99,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Concat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['train_or_test']='train'\n",
    "test['train_or_test']='test'\n",
    "df=pd.concat([train,test])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### numeric aggregate feataures grouped by patient_ID"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "def agg_numeric(df, parent_var, df_name):\n",
    "    \"\"\"\n",
    "    Groups and aggregates the numeric values in a child dataframe\n",
    "    by the parent variable.\n",
    "    \n",
    "    Parameters\n",
    "    --------\n",
    "        df (dataframe): \n",
    "            the child dataframe to calculate the statistics on\n",
    "        parent_var (string): \n",
    "            the parent variable used for grouping and aggregating\n",
    "        df_name (string): \n",
    "            the variable used to rename the columns\n",
    "        \n",
    "    Return\n",
    "    --------\n",
    "        agg (dataframe): \n",
    "            a dataframe with the statistics aggregated by the `parent_var` for \n",
    "            all numeric columns. Each observation of the parent variable will have \n",
    "            one row in the dataframe with the parent variable as the index. \n",
    "            The columns are also renamed using the `df_name`. Columns with all duplicate\n",
    "            values are removed. \n",
    "    \n",
    "    \"\"\"\n",
    "    \n",
    "            \n",
    "    # Only want the numeric variables\n",
    "    parent_ids = df[parent_var].copy()\n",
    "    numeric_df = df.select_dtypes('number').drop(columns={'Patient_ID', 'Health_Camp_ID','target'}).copy()\n",
    "    numeric_df[parent_var] = parent_ids\n",
    "\n",
    "    # Group by the specified variable and calculate the statistics\n",
    "    agg = numeric_df.groupby(parent_var).agg(['count', 'mean', 'max', 'min', 'sum'])\n",
    "\n",
    "    # Need to create new column names\n",
    "    columns = []\n",
    "\n",
    "    # Iterate through the variables names\n",
    "    for var in agg.columns.levels[0]:\n",
    "        if var != parent_var:\n",
    "            # Iterate through the stat names\n",
    "            for stat in agg.columns.levels[1]:\n",
    "                # Make a new column name for the variable and stat\n",
    "                columns.append('%s_%s_%s' % (df_name, var, stat))\n",
    "    \n",
    "    agg.columns = columns\n",
    "    \n",
    "    # Remove the columns with all redundant values\n",
    "    _, idx = np.unique(agg, axis = 1, return_index=True)\n",
    "    agg = agg.iloc[:, idx]\n",
    "    \n",
    "    return agg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "PID aggregate shape:  (36692, 141)\n"
     ]
    }
   ],
   "source": [
    "PID_aggregate = agg_numeric(df, 'Patient_ID', 'agg')\n",
    "print('PID aggregate shape: ', PID_aggregate.shape)\n",
    "df=df.merge(PID_aggregate, on ='Patient_ID', how = 'left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['Patient_Registered_before_days']=pd.to_datetime(df['Camp_Start_Date'])-pd.to_datetime(df['Registration_Date'])\n",
    "df['Patient_Registered_before_days']=df['Patient_Registered_before_days'].apply(timediff)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "train=df.loc[df.train_or_test.isin(['train'])]\n",
    "test=df.loc[df.train_or_test.isin(['test'])]\n",
    "train.drop(columns={'train_or_test'},axis=1,inplace=True)\n",
    "test.drop(columns={'train_or_test'},axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Creating Time based validation for 5 months ,as same as public and private test period"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "trn=train[train['Camp_Start_Date'] <'2005-11-01']\n",
    "val=train[train['Camp_Start_Date'] >'2005-10-30']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Age', 'Camp_Duration', 'Camp_End_Date', 'Camp_Start_Date', 'Category1',\n",
       "       'Category2', 'Category3', 'City_Type', 'Education_Score',\n",
       "       'Employer_Category',\n",
       "       ...\n",
       "       'agg_Income_mean', 'agg_Income_max', 'agg_Income_min', 'agg_Age_mean',\n",
       "       'agg_Age_max', 'agg_Age_min', 'agg_Education_Score_mean',\n",
       "       'agg_Education_Score_max', 'agg_Education_Score_min',\n",
       "       'Patient_Registered_before_days'],\n",
       "      dtype='object', length=187)"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "178"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TARGET_COL = 'target'\n",
    "features = [c for c in trn.columns if c not in ['Patient_ID', 'Health_Camp_ID','Category3','Registration_Date', 'id','target','Camp_Start_Date','Camp_End_Date','First_Interaction',TARGET_COL]]\n",
    "len(features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### validating the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "from lightgbm import LGBMClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 100 rounds\n",
      "[50]\tvalid_0's auc: 0.836279\tvalid_0's binary_logloss: 0.425735\n",
      "[100]\tvalid_0's auc: 0.840514\tvalid_0's binary_logloss: 0.402239\n",
      "[150]\tvalid_0's auc: 0.844641\tvalid_0's binary_logloss: 0.394518\n",
      "[200]\tvalid_0's auc: 0.846119\tvalid_0's binary_logloss: 0.392112\n",
      "[250]\tvalid_0's auc: 0.846637\tvalid_0's binary_logloss: 0.390968\n",
      "[300]\tvalid_0's auc: 0.846511\tvalid_0's binary_logloss: 0.390132\n",
      "[350]\tvalid_0's auc: 0.846273\tvalid_0's binary_logloss: 0.389972\n",
      "Early stopping, best iteration is:\n",
      "[276]\tvalid_0's auc: 0.847016\tvalid_0's binary_logloss: 0.390273\n"
     ]
    }
   ],
   "source": [
    "clf = LGBMClassifier(n_estimators=550,\n",
    "                     learning_rate=0.03,\n",
    "                     min_child_samples=40,\n",
    "                     random_state=1,\n",
    "                     colsample_bytree=0.5,\n",
    "                     reg_alpha=2,\n",
    "                     reg_lambda=2)\n",
    "\n",
    "clf.fit(trn[features], trn[TARGET_COL], eval_set=[(val[features], val[TARGET_COL])], verbose=50,\n",
    "        eval_metric = 'auc', early_stopping_rounds = 100)\n",
    "\n",
    "preds = clf.predict_proba(test[features])[:, 1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Removing 0 importance feature didnt help"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x20d80e85748>"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fi = pd.Series(index = features, data = clf.feature_importances_)\n",
    "fi.sort_values(ascending=False)[0:20][::-1].plot(kind = 'barh')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Final Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Blending with different seeds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# preds = 0\n",
    "# for seed_val, num_rounds in [[1,450], [1997,450], [2020, 450]]:\n",
    "#     print (seed_val, num_rounds)\n",
    "#     m=LGBMClassifier(n_estimators=num_rounds,learning_rate=0.03,random_state=seed_val,colsample_bytree=0.5,reg_alpha=2,reg_lambda=2)\n",
    "#     m.fit(train[features],train['target'])\n",
    "#     predict=m.predict_proba(test[features])[:,1]\n",
    "#     preds += predict\n",
    "# preds = preds/3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "3\n",
      "10\n",
      "15\n",
      "20\n",
      "33\n",
      "333\n",
      "1997\n",
      "2020\n",
      "2021\n"
     ]
    }
   ],
   "source": [
    "preds = 0\n",
    "for seed_val in [1,3,10,15,20,33,333,1997,2020,2021]:\n",
    "    print (seed_val)\n",
    "    m=LGBMClassifier(n_estimators=450,learning_rate=0.03,random_state=seed_val,colsample_bytree=0.5,reg_alpha=2,reg_lambda=2)\n",
    "    m.fit(train[features],train['target'])\n",
    "    predict=m.predict_proba(test[features])[:,1]\n",
    "    preds += predict\n",
    "preds = preds/10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub = pd.DataFrame({\"Patient_ID\":test.Patient_ID.values})\n",
    "sub[\"Health_Camp_ID\"] = test.Health_Camp_ID.values\n",
    "sub[\"Outcome\"] =  preds\n",
    "sub.to_csv(\"lgbmblending.csv\", index=False)"
   ]
  },
  {
   "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}