SVC using square root transformation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████████████████████| 2/2 [20:43<00:00, 621.62s/it]\n"
     ]
    },
    {
     "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>Model</th>\n",
       "      <th>Accuracy Mean</th>\n",
       "      <th>Accuracy Std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>SVC with reciprocal trans</td>\n",
       "      <td>0.807512</td>\n",
       "      <td>0.009886</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>SVC no trans</td>\n",
       "      <td>0.803100</td>\n",
       "      <td>0.012257</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       Model  Accuracy Mean  Accuracy Std\n",
       "1  SVC with reciprocal trans       0.807512      0.009886\n",
       "0               SVC no trans       0.803100      0.012257"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Importing libraries\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "\n",
    "# Load the data\n",
    "df = pd.read_csv('credit_risk.csv')\n",
    "\n",
    "# Importing liberaries for preprocessing\n",
    "from sklearn.impute import KNNImputer\n",
    "from sklearn.preprocessing import OneHotEncoder, MinMaxScaler, RobustScaler, MaxAbsScaler\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.pipeline import Pipeline\n",
    "\n",
    "# Drop Id column\n",
    "df = df.drop(['Id'], axis=1)\n",
    "\n",
    "Scale_cols = ['Age','Income','Amount','Percent_income','Cred_length']\n",
    "impute_cols = ['Emp_length', 'Rate']\n",
    "encode_cols = df.select_dtypes(exclude='number').columns.to_list()\n",
    "\n",
    "# logarithmic transformation in the features\n",
    "df_log = df.copy()\n",
    "for col in Scale_cols:\n",
    "    df_log[col] = df_log[col] + (df[df[col]!=0][col].min()/2)\n",
    "    df_log[col] = np.sqrt(df_log[col])\n",
    "    \n",
    "# Impute missing values and scaling\n",
    "imputer_mean= Pipeline([\n",
    "    ('imputer', KNNImputer(n_neighbors=5)),\n",
    "    ('scaling', RobustScaler())\n",
    "])\n",
    "\n",
    "Preprocessing= ColumnTransformer(\n",
    "    transformers=[\n",
    "        ('imputer', imputer_mean, impute_cols),\n",
    "        ('encoder', OneHotEncoder(handle_unknown='ignore'), encode_cols)\n",
    "    ], remainder=\"passthrough\"\n",
    ")\n",
    "\n",
    "\n",
    "# Classificatin models\n",
    "from sklearn.svm import SVC\n",
    "\n",
    "# Model processing\n",
    "from sklearn.model_selection import train_test_split, KFold, cross_val_score\n",
    "from tqdm import tqdm\n",
    "\n",
    "# Split x and y for the model\n",
    "y = df['Status']\n",
    "x = df.drop(['Status'], axis=1)\n",
    "\n",
    "# Split x and y for the model\n",
    "y_log = df_log['Status']\n",
    "x_log = df_log.drop(['Status'], axis=1)\n",
    "\n",
    "# Split data to Train and Test data\n",
    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=24, stratify=y)\n",
    "X_train_log, X_test_log, y_train_log, y_test_log = train_test_split(x_log, y_log, test_size=0.2, random_state=24, stratify=y)\n",
    "\n",
    "model = Pipeline([\n",
    "        ('preprocessing', Preprocessing),\n",
    "        ('model', SVC())\n",
    "    ])\n",
    "\n",
    "# The data\n",
    "data = [[X_train, y_train], [X_train_log, y_train_log]]\n",
    "\n",
    "# List for score matrix\n",
    "accuracy_score = []\n",
    "accuracy_mean = []\n",
    "accuracy_std = []\n",
    "\n",
    "# Calculate model's matrix score\n",
    "for feat in tqdm(data):\n",
    "    \n",
    "    # Cross validate split\n",
    "    crossval = KFold(n_splits=5, shuffle=True, random_state=24)\n",
    "\n",
    "    # Calculate the accuracy from cross validation\n",
    "    accuracy = cross_val_score(\n",
    "        model, \n",
    "        feat[0], \n",
    "        feat[1], \n",
    "        cv=crossval, \n",
    "        scoring='accuracy', \n",
    "        error_score='raise'\n",
    "        )\n",
    "\n",
    "    #accuracy_score.append(accuracy)\n",
    "    accuracy_mean.append(accuracy.mean())\n",
    "    accuracy_std.append(accuracy.std())\n",
    "\n",
    "# Model Matrix Evaluation\n",
    "model_matrix = pd.DataFrame({\n",
    "    'Model': ['SVC no trans', 'SVC with reciprocal trans'],\n",
    "    'Accuracy Mean': accuracy_mean,\n",
    "    'Accuracy Std': accuracy_std\n",
    "})\n",
    "model_matrix.sort_values(by='Accuracy Mean', ascending=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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}