03.B. Model Development - Fraud After Transaction.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import xgboost as xgb\n",
    "from xgboost import plot_importance\n",
    "from xgboost.sklearn import XGBClassifier\n",
    "\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import cross_validate, learning_curve, GridSearchCV\n",
    "from sklearn.metrics import average_precision_score, roc_curve, precision_recall_curve, confusion_matrix\n",
    "\n",
    "import matplotlib.pylab as plt\n",
    "%matplotlib inline\n",
    "from matplotlib.pylab import rcParams\n",
    "rcParams['figure.figsize'] = 12, 4\n",
    "\n",
    "import time\n",
    "import pickle"
   ]
  },
  {
   "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>step</th>\n",
       "      <th>type</th>\n",
       "      <th>amount</th>\n",
       "      <th>nameOrig</th>\n",
       "      <th>oldBalanceOrig</th>\n",
       "      <th>newBalanceOrig</th>\n",
       "      <th>nameDest</th>\n",
       "      <th>oldBalanceDest</th>\n",
       "      <th>newBalanceDest</th>\n",
       "      <th>isFraud</th>\n",
       "      <th>...</th>\n",
       "      <th>time_of_day</th>\n",
       "      <th>amount_tod_median</th>\n",
       "      <th>amount_tod_std</th>\n",
       "      <th>amount_tod_max</th>\n",
       "      <th>amount_deviance_5</th>\n",
       "      <th>amount_deviance_6</th>\n",
       "      <th>amount_deviance_7</th>\n",
       "      <th>amount_deviance_8</th>\n",
       "      <th>diffBalanceOrig</th>\n",
       "      <th>diffBalanceDest</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>181.00</td>\n",
       "      <td>C1305486145</td>\n",
       "      <td>181.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C553264065</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>156986.25</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>-0.42</td>\n",
       "      <td>-156805.25</td>\n",
       "      <td>4407703.05</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>181.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>CASH_OUT</td>\n",
       "      <td>181.00</td>\n",
       "      <td>C840083671</td>\n",
       "      <td>181.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C38997010</td>\n",
       "      <td>21182.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>156986.25</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>-0.42</td>\n",
       "      <td>-156805.25</td>\n",
       "      <td>4407703.05</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>21363.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>CASH_OUT</td>\n",
       "      <td>229133.94</td>\n",
       "      <td>C905080434</td>\n",
       "      <td>15325.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C476402209</td>\n",
       "      <td>5083.0</td>\n",
       "      <td>51513.44</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>156986.25</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.19</td>\n",
       "      <td>72147.69</td>\n",
       "      <td>4178750.11</td>\n",
       "      <td>0.95</td>\n",
       "      <td>213808.94</td>\n",
       "      <td>182703.5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>215310.30</td>\n",
       "      <td>C1670993182</td>\n",
       "      <td>705.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C1100439041</td>\n",
       "      <td>22425.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>156986.25</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.16</td>\n",
       "      <td>58324.05</td>\n",
       "      <td>4192573.75</td>\n",
       "      <td>0.95</td>\n",
       "      <td>214605.30</td>\n",
       "      <td>237735.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>311685.89</td>\n",
       "      <td>C1984094095</td>\n",
       "      <td>10835.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C932583850</td>\n",
       "      <td>6267.0</td>\n",
       "      <td>2719172.89</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>156986.25</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.41</td>\n",
       "      <td>154699.64</td>\n",
       "      <td>4096198.16</td>\n",
       "      <td>0.93</td>\n",
       "      <td>300850.89</td>\n",
       "      <td>-2401220.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 35 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   step      type     amount     nameOrig  oldBalanceOrig  newBalanceOrig  \\\n",
       "0     1  TRANSFER     181.00  C1305486145           181.0             0.0   \n",
       "1     1  CASH_OUT     181.00   C840083671           181.0             0.0   \n",
       "2     1  CASH_OUT  229133.94   C905080434         15325.0             0.0   \n",
       "3     1  TRANSFER  215310.30  C1670993182           705.0             0.0   \n",
       "4     1  TRANSFER  311685.89  C1984094095         10835.0             0.0   \n",
       "\n",
       "      nameDest  oldBalanceDest  newBalanceDest  isFraud  ...  time_of_day  \\\n",
       "0   C553264065             0.0            0.00        1  ...            1   \n",
       "1    C38997010         21182.0            0.00        1  ...            1   \n",
       "2   C476402209          5083.0        51513.44        0  ...            1   \n",
       "3  C1100439041         22425.0            0.00        0  ...            1   \n",
       "4   C932583850          6267.0      2719172.89        0  ...            1   \n",
       "\n",
       "   amount_tod_median  amount_tod_std  amount_tod_max  amount_deviance_5  \\\n",
       "0          156986.25       375750.01      4407884.05              -0.42   \n",
       "1          156986.25       375750.01      4407884.05              -0.42   \n",
       "2          156986.25       375750.01      4407884.05               0.19   \n",
       "3          156986.25       375750.01      4407884.05               0.16   \n",
       "4          156986.25       375750.01      4407884.05               0.41   \n",
       "\n",
       "   amount_deviance_6  amount_deviance_7  amount_deviance_8  diffBalanceOrig  \\\n",
       "0         -156805.25         4407703.05               1.00             0.00   \n",
       "1         -156805.25         4407703.05               1.00             0.00   \n",
       "2           72147.69         4178750.11               0.95        213808.94   \n",
       "3           58324.05         4192573.75               0.95        214605.30   \n",
       "4          154699.64         4096198.16               0.93        300850.89   \n",
       "\n",
       "   diffBalanceDest  \n",
       "0            181.0  \n",
       "1          21363.0  \n",
       "2         182703.5  \n",
       "3         237735.3  \n",
       "4       -2401220.0  \n",
       "\n",
       "[5 rows x 35 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_cot=pd.read_csv('/Users/Abhirup/Desktop/Projects/HSBC DS Assignment/df_cot.csv')\n",
    "df_cot.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2770409, 35)\n",
      "step                   int64\n",
      "type                  object\n",
      "amount               float64\n",
      "nameOrig              object\n",
      "oldBalanceOrig       float64\n",
      "newBalanceOrig       float64\n",
      "nameDest              object\n",
      "oldBalanceDest       float64\n",
      "newBalanceDest       float64\n",
      "isFraud                int64\n",
      "isFlaggedFraud         int64\n",
      "week                   int64\n",
      "$TransLW             float64\n",
      "NumTransLW             int64\n",
      "FraudBefore            int64\n",
      "oldBalanceOrigCat      int64\n",
      "amount_median        float64\n",
      "amount_std           float64\n",
      "amt_max              float64\n",
      "amount_deviance_1    float64\n",
      "amount_deviance_2    float64\n",
      "amount_deviance_3    float64\n",
      "amount_deviance_4    float64\n",
      "amount_zero            int64\n",
      "amt_bal_perc         float64\n",
      "time_of_day            int64\n",
      "amount_tod_median    float64\n",
      "amount_tod_std       float64\n",
      "amount_tod_max       float64\n",
      "amount_deviance_5    float64\n",
      "amount_deviance_6    float64\n",
      "amount_deviance_7    float64\n",
      "amount_deviance_8    float64\n",
      "diffBalanceOrig      float64\n",
      "diffBalanceDest      float64\n",
      "dtype: object\n"
     ]
    }
   ],
   "source": [
    "print(df_cot.shape)\n",
    "print(df_cot.dtypes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    2237500\n",
       "1     532909\n",
       "Name: trans_type, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_cot.loc[df_cot['type']=='TRANSFER','trans_type']=1\n",
    "df_cot.loc[df_cot['type']=='CASH_OUT','trans_type']=0\n",
    "df_cot['trans_type']=df_cot['trans_type'].astype(int)\n",
    "df_cot['trans_type'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>step</th>\n",
       "      <th>type</th>\n",
       "      <th>amount</th>\n",
       "      <th>nameOrig</th>\n",
       "      <th>oldBalanceOrig</th>\n",
       "      <th>newBalanceOrig</th>\n",
       "      <th>nameDest</th>\n",
       "      <th>oldBalanceDest</th>\n",
       "      <th>newBalanceDest</th>\n",
       "      <th>isFraud</th>\n",
       "      <th>...</th>\n",
       "      <th>amount_tod_std</th>\n",
       "      <th>amount_tod_max</th>\n",
       "      <th>amount_deviance_5</th>\n",
       "      <th>amount_deviance_6</th>\n",
       "      <th>amount_deviance_7</th>\n",
       "      <th>amount_deviance_8</th>\n",
       "      <th>diffBalanceOrig</th>\n",
       "      <th>diffBalanceDest</th>\n",
       "      <th>trans_type</th>\n",
       "      <th>tran_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>181.00</td>\n",
       "      <td>C1305486145</td>\n",
       "      <td>181.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C553264065</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>-0.42</td>\n",
       "      <td>-156805.25</td>\n",
       "      <td>4407703.05</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>181.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>CASH_OUT</td>\n",
       "      <td>181.00</td>\n",
       "      <td>C840083671</td>\n",
       "      <td>181.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C38997010</td>\n",
       "      <td>21182.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>-0.42</td>\n",
       "      <td>-156805.25</td>\n",
       "      <td>4407703.05</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>21363.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>CASH_OUT</td>\n",
       "      <td>229133.94</td>\n",
       "      <td>C905080434</td>\n",
       "      <td>15325.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C476402209</td>\n",
       "      <td>5083.0</td>\n",
       "      <td>51513.44</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.19</td>\n",
       "      <td>72147.69</td>\n",
       "      <td>4178750.11</td>\n",
       "      <td>0.95</td>\n",
       "      <td>213808.94</td>\n",
       "      <td>182703.5</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>215310.30</td>\n",
       "      <td>C1670993182</td>\n",
       "      <td>705.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C1100439041</td>\n",
       "      <td>22425.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.16</td>\n",
       "      <td>58324.05</td>\n",
       "      <td>4192573.75</td>\n",
       "      <td>0.95</td>\n",
       "      <td>214605.30</td>\n",
       "      <td>237735.3</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>TRANSFER</td>\n",
       "      <td>311685.89</td>\n",
       "      <td>C1984094095</td>\n",
       "      <td>10835.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>C932583850</td>\n",
       "      <td>6267.0</td>\n",
       "      <td>2719172.89</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>375750.01</td>\n",
       "      <td>4407884.05</td>\n",
       "      <td>0.41</td>\n",
       "      <td>154699.64</td>\n",
       "      <td>4096198.16</td>\n",
       "      <td>0.93</td>\n",
       "      <td>300850.89</td>\n",
       "      <td>-2401220.0</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 37 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   step      type     amount     nameOrig  oldBalanceOrig  newBalanceOrig  \\\n",
       "0     1  TRANSFER     181.00  C1305486145           181.0             0.0   \n",
       "1     1  CASH_OUT     181.00   C840083671           181.0             0.0   \n",
       "2     1  CASH_OUT  229133.94   C905080434         15325.0             0.0   \n",
       "3     1  TRANSFER  215310.30  C1670993182           705.0             0.0   \n",
       "4     1  TRANSFER  311685.89  C1984094095         10835.0             0.0   \n",
       "\n",
       "      nameDest  oldBalanceDest  newBalanceDest  isFraud  ...  amount_tod_std  \\\n",
       "0   C553264065             0.0            0.00        1  ...       375750.01   \n",
       "1    C38997010         21182.0            0.00        1  ...       375750.01   \n",
       "2   C476402209          5083.0        51513.44        0  ...       375750.01   \n",
       "3  C1100439041         22425.0            0.00        0  ...       375750.01   \n",
       "4   C932583850          6267.0      2719172.89        0  ...       375750.01   \n",
       "\n",
       "   amount_tod_max  amount_deviance_5  amount_deviance_6  amount_deviance_7  \\\n",
       "0      4407884.05              -0.42         -156805.25         4407703.05   \n",
       "1      4407884.05              -0.42         -156805.25         4407703.05   \n",
       "2      4407884.05               0.19           72147.69         4178750.11   \n",
       "3      4407884.05               0.16           58324.05         4192573.75   \n",
       "4      4407884.05               0.41          154699.64         4096198.16   \n",
       "\n",
       "   amount_deviance_8  diffBalanceOrig  diffBalanceDest  trans_type  tran_id  \n",
       "0               1.00             0.00            181.0           1        1  \n",
       "1               1.00             0.00          21363.0           0        2  \n",
       "2               0.95        213808.94         182703.5           0        3  \n",
       "3               0.95        214605.30         237735.3           1        4  \n",
       "4               0.93        300850.89       -2401220.0           1        5  \n",
       "\n",
       "[5 rows x 37 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Creating a transaction ID\n",
    "df_cot['tran_id']=df_cot.index +1\n",
    "df_cot.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# not including new balances assuming it gets updated post transaction\n",
    "feature_list=['NumTransLW','FraudBefore','oldBalanceOrigCat','amount_zero','time_of_day','amount','oldBalanceOrig'\n",
    "              ,'oldBalanceDest','$TransLW','amount_median','amount_std','amt_max','amount_deviance_1'\n",
    "              ,'amount_deviance_2','amount_deviance_3','amount_deviance_4','amt_bal_perc','amount_tod_median'\n",
    "              ,'amount_tod_std','amount_tod_max','amount_deviance_5','amount_deviance_6','amount_deviance_7'\n",
    "              ,'amount_deviance_8','diffBalanceOrig','diffBalanceDest','trans_type']\n",
    "variable_list=['tran_id','isFraud','NumTransLW','FraudBefore','oldBalanceOrigCat','amount_zero','time_of_day','amount'\n",
    "               ,'oldBalanceOrig','oldBalanceDest','$TransLW','amount_median','amount_std','amt_max'\n",
    "               ,'amount_deviance_1','amount_deviance_2','amount_deviance_3','amount_deviance_4','amt_bal_perc'\n",
    "               ,'amount_tod_median','amount_tod_std','amount_tod_max','amount_deviance_5','amount_deviance_6'\n",
    "               ,'amount_deviance_7','amount_deviance_8','diffBalanceOrig','diffBalanceDest','trans_type']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1697805, 29)\n",
      "(81630, 29)\n",
      "0    1693762\n",
      "1       4043\n",
      "Name: isFraud, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "df_model_dev=df_cot.loc[df_cot['week'].isin([2,3]),variable_list]\n",
    "df_oot_val=df_cot.loc[df_cot['week']==4,variable_list]\n",
    "print(df_model_dev.shape)\n",
    "print(df_oot_val.shape)\n",
    "print(df_model_dev['isFraud'].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "target='isFraud'\n",
    "IDcol='tran_id'\n",
    "\n",
    "label=df_model_dev[target]\n",
    "\n",
    "#split data into train and test\n",
    "seed=1234\n",
    "test_size=0.3\n",
    "\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(df_model_dev[feature_list], \n",
    "                                                    label, test_size=test_size, random_state=seed,stratify=label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "42.86\n",
      "43 2\n"
     ]
    }
   ],
   "source": [
    "print(round(Y_test.sum()/Y_train.sum()*100,2))\n",
    "print(round(len(X_test)/len(X_train)*100),2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/anaconda3/lib/python3.7/site-packages/xgboost/core.py:587: FutureWarning: Series.base is deprecated and will be removed in a future version\n",
      "  if getattr(data, 'base', None) is not None and \\\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "198.75746181805928\n"
     ]
    }
   ],
   "source": [
    "xgb_vs_1 = XGBClassifier(\n",
    " learning_rate =0.05,\n",
    " n_estimators=1300,\n",
    " max_depth=5,\n",
    " min_child_weight=1,\n",
    " gamma=0,\n",
    " subsample=0.8,\n",
    " colsample_bytree=0.8,\n",
    " objective= 'binary:logistic',\n",
    " nthread=4,\n",
    " scale_pos_weight=419,\n",
    " seed=27)\n",
    "\n",
    "train_dm=xgb.DMatrix(X_train.values, label=Y_train)\n",
    "\n",
    "start_time=time.time()\n",
    "\n",
    "cvresult=xgb.cv(xgb_vs_1.get_xgb_params(), train_dm, num_boost_round=xgb_vs_1.get_params()['n_estimators'], nfold=5,\n",
    "                metrics='aucpr', early_stopping_rounds=50)\n",
    "\n",
    "print((time.time()-start_time)/60)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "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>train-aucpr-mean</th>\n",
       "      <th>train-aucpr-std</th>\n",
       "      <th>test-aucpr-mean</th>\n",
       "      <th>test-aucpr-std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>1164</td>\n",
       "      <td>0.999977</td>\n",
       "      <td>0.000014</td>\n",
       "      <td>0.991951</td>\n",
       "      <td>0.002078</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1165</td>\n",
       "      <td>0.999977</td>\n",
       "      <td>0.000014</td>\n",
       "      <td>0.991961</td>\n",
       "      <td>0.002079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1166</td>\n",
       "      <td>0.999977</td>\n",
       "      <td>0.000014</td>\n",
       "      <td>0.991954</td>\n",
       "      <td>0.002081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1167</td>\n",
       "      <td>0.999977</td>\n",
       "      <td>0.000013</td>\n",
       "      <td>0.991970</td>\n",
       "      <td>0.002079</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1168</td>\n",
       "      <td>0.999977</td>\n",
       "      <td>0.000013</td>\n",
       "      <td>0.991975</td>\n",
       "      <td>0.002071</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      train-aucpr-mean  train-aucpr-std  test-aucpr-mean  test-aucpr-std\n",
       "1164          0.999977         0.000014         0.991951        0.002078\n",
       "1165          0.999977         0.000014         0.991961        0.002079\n",
       "1166          0.999977         0.000014         0.991954        0.002081\n",
       "1167          0.999977         0.000013         0.991970        0.002079\n",
       "1168          0.999977         0.000013         0.991975        0.002071"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    " cvresult.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "      train-aucpr-mean  train-aucpr-std  test-aucpr-mean  test-aucpr-std\n",
      "1168          0.999977         0.000013         0.991975        0.002071\n",
      "      train-aucpr-mean  train-aucpr-std  test-aucpr-mean  test-aucpr-std\n",
      "1168          0.999977         0.000013         0.991975        0.002071\n"
     ]
    }
   ],
   "source": [
    "print(cvresult[cvresult['test-aucpr-mean'] == cvresult['test-aucpr-mean'].max()])\n",
    "print(cvresult[cvresult['train-aucpr-mean'] == cvresult['train-aucpr-mean'].max()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1169"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cvresult.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_vs_f = XGBClassifier(\n",
    " learning_rate =0.05,\n",
    " n_estimators=1169,\n",
    " max_depth=5,\n",
    " min_child_weight=1,\n",
    " gamma=0,\n",
    " subsample=0.8,\n",
    " colsample_bytree=0.8,\n",
    " objective= 'binary:logistic',\n",
    " nthread=4,\n",
    " scale_pos_weight=419,\n",
    " seed=27)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1188463, 29)\n",
      "(509342, 29)\n"
     ]
    }
   ],
   "source": [
    "train_df=df_model_dev.loc[X_train.index,:]\n",
    "test_df=df_model_dev.loc[X_test.index,:]\n",
    "print(train_df.shape)\n",
    "print(test_df.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/Users/Abhirup/Desktop/Projects/HSBC DS Assignment'"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "57.78119538227717\n"
     ]
    }
   ],
   "source": [
    "start_time=time.time()\n",
    "xgb_vs_f.fit(train_df[feature_list], train_df[target],eval_metric= ['auc', 'aucpr'])\n",
    "pickle.dump(xgb_vs_f, open(\"xgb_model_2\", \"wb\"))\n",
    "print((time.time()-start_time)/60)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUPRC Train= 0.9999452958516887\n",
      "AUPRC Test= 0.992975889269172\n"
     ]
    }
   ],
   "source": [
    "train_df['Prob_Score']=xgb_vs_f.predict_proba(train_df[feature_list])[:,1]\n",
    "test_df['Prob_Score']=xgb_vs_f.predict_proba(test_df[feature_list])[:,1]\n",
    "print('AUPRC Train= {}'.format(average_precision_score(train_df[target], train_df['Prob_Score'])))\n",
    "print('AUPRC Test= {}'.format(average_precision_score(test_df[target], test_df['Prob_Score'])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Check Feature Importance\n",
    "plot_importance(xgb_vs_f, max_num_features=15)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUPRC OOT= 0.997849120785378\n"
     ]
    }
   ],
   "source": [
    "df_oot_val['Prob_Score']=xgb_vs_f.predict_proba(df_oot_val[feature_list])[:,1]\n",
    "print('AUPRC OOT= {}'.format(average_precision_score(df_oot_val[target], df_oot_val['Prob_Score'])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "precision, recall, thresholds = precision_recall_curve(train_df[target], train_df['Prob_Score'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9991823"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Transaction Score\n",
    "thresholds[np.argmax(precision)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "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>tran_id</th>\n",
       "      <th>isFraud</th>\n",
       "      <th>NumTransLW</th>\n",
       "      <th>FraudBefore</th>\n",
       "      <th>oldBalanceOrigCat</th>\n",
       "      <th>amount_zero</th>\n",
       "      <th>time_of_day</th>\n",
       "      <th>amount</th>\n",
       "      <th>oldBalanceOrig</th>\n",
       "      <th>oldBalanceDest</th>\n",
       "      <th>...</th>\n",
       "      <th>amount_tod_max</th>\n",
       "      <th>amount_deviance_5</th>\n",
       "      <th>amount_deviance_6</th>\n",
       "      <th>amount_deviance_7</th>\n",
       "      <th>amount_deviance_8</th>\n",
       "      <th>diffBalanceOrig</th>\n",
       "      <th>diffBalanceDest</th>\n",
       "      <th>trans_type</th>\n",
       "      <th>Prob_Score</th>\n",
       "      <th>isFraud_Pred</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>2486318</td>\n",
       "      <td>2486319</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>140499.38</td>\n",
       "      <td>10575.0</td>\n",
       "      <td>620769.42</td>\n",
       "      <td>...</td>\n",
       "      <td>47504216.38</td>\n",
       "      <td>-0.09</td>\n",
       "      <td>-61928.34</td>\n",
       "      <td>47363717.00</td>\n",
       "      <td>1.0</td>\n",
       "      <td>129924.38</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>4.755995e-08</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2212281</td>\n",
       "      <td>2212282</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>54612.50</td>\n",
       "      <td>0.0</td>\n",
       "      <td>884088.71</td>\n",
       "      <td>...</td>\n",
       "      <td>69886731.30</td>\n",
       "      <td>-0.12</td>\n",
       "      <td>-109859.74</td>\n",
       "      <td>69832118.80</td>\n",
       "      <td>1.0</td>\n",
       "      <td>54612.50</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2.895246e-07</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2121081</td>\n",
       "      <td>2121082</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>311824.10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2088728.46</td>\n",
       "      <td>...</td>\n",
       "      <td>92445516.64</td>\n",
       "      <td>0.12</td>\n",
       "      <td>140882.19</td>\n",
       "      <td>92133692.54</td>\n",
       "      <td>1.0</td>\n",
       "      <td>311824.10</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>7.974330e-07</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1977064</td>\n",
       "      <td>1977065</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>33010.29</td>\n",
       "      <td>39725.0</td>\n",
       "      <td>303816.90</td>\n",
       "      <td>...</td>\n",
       "      <td>57787800.93</td>\n",
       "      <td>-0.21</td>\n",
       "      <td>-171613.69</td>\n",
       "      <td>57754790.64</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>4.410386e-06</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1298699</td>\n",
       "      <td>1298700</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>150102.72</td>\n",
       "      <td>0.0</td>\n",
       "      <td>969143.65</td>\n",
       "      <td>...</td>\n",
       "      <td>92445516.64</td>\n",
       "      <td>-0.02</td>\n",
       "      <td>-20839.19</td>\n",
       "      <td>92295413.92</td>\n",
       "      <td>1.0</td>\n",
       "      <td>150102.72</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>4.538856e-07</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         tran_id  isFraud  NumTransLW  FraudBefore  oldBalanceOrigCat  \\\n",
       "2486318  2486319        0           0            0                  2   \n",
       "2212281  2212282        0           0            0                  1   \n",
       "2121081  2121082        0           0            0                  1   \n",
       "1977064  1977065        0           0            0                  3   \n",
       "1298699  1298700        0           0            0                  1   \n",
       "\n",
       "         amount_zero  time_of_day     amount  oldBalanceOrig  oldBalanceDest  \\\n",
       "2486318            0            2  140499.38         10575.0       620769.42   \n",
       "2212281            0            3   54612.50             0.0       884088.71   \n",
       "2121081            0            2  311824.10             0.0      2088728.46   \n",
       "1977064            0            2   33010.29         39725.0       303816.90   \n",
       "1298699            0            2  150102.72             0.0       969143.65   \n",
       "\n",
       "         ...  amount_tod_max  amount_deviance_5  amount_deviance_6  \\\n",
       "2486318  ...     47504216.38              -0.09          -61928.34   \n",
       "2212281  ...     69886731.30              -0.12         -109859.74   \n",
       "2121081  ...     92445516.64               0.12          140882.19   \n",
       "1977064  ...     57787800.93              -0.21         -171613.69   \n",
       "1298699  ...     92445516.64              -0.02          -20839.19   \n",
       "\n",
       "         amount_deviance_7  amount_deviance_8  diffBalanceOrig  \\\n",
       "2486318        47363717.00                1.0        129924.38   \n",
       "2212281        69832118.80                1.0         54612.50   \n",
       "2121081        92133692.54                1.0        311824.10   \n",
       "1977064        57754790.64                1.0             0.00   \n",
       "1298699        92295413.92                1.0        150102.72   \n",
       "\n",
       "         diffBalanceDest  trans_type    Prob_Score  isFraud_Pred  \n",
       "2486318              0.0           0  4.755995e-08             0  \n",
       "2212281              0.0           0  2.895246e-07             0  \n",
       "2121081              0.0           0  7.974330e-07             0  \n",
       "1977064              0.0           0  4.410386e-06             0  \n",
       "1298699              0.0           0  4.538856e-07             0  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.loc[train_df['Prob_Score']>=thresholds[np.argmax(precision)],'isFraud_Pred']=1\n",
    "train_df.loc[train_df['Prob_Score']<thresholds[np.argmax(precision)],'isFraud_Pred']=0\n",
    "train_df['isFraud_Pred']=train_df['isFraud_Pred'].astype(int)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      " [[1185633       0]\n",
      " [    149    2681]]\n",
      "\n",
      "Accuracy: 99.99\n",
      "\n",
      "Sensitivity/Recall/TPR: 94.73\n",
      "\n",
      "Precision: 100.0\n",
      "\n",
      "Specificity:\n"
     ]
    }
   ],
   "source": [
    "cm1=confusion_matrix(train_df[target],train_df['isFraud_Pred'])\n",
    "print('Confusion Matrix: \\n',cm1)\n",
    "total=sum(sum(cm1))\n",
    "print(\"\\nAccuracy: {}\" .format(round(((cm1[0,0]+cm1[1,1])/total)*100,2)))\n",
    "print(\"\\nSensitivity/Recall/TPR: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[1,0]))*100,2)))\n",
    "print(\"\\nPrecision: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[0,1]))*100,2)))\n",
    "print(\"\\nSpecificity:\" .format(round((cm1[0,0]/(cm1[0,0]+cm1[0,1]))*100,2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      " [[508129      0]\n",
      " [   107   1106]]\n",
      "\n",
      "Accuracy: 99.98\n",
      "\n",
      "Sensitivity/Recall/TPR: 91.18\n",
      "\n",
      "Precision: 100.0\n",
      "\n",
      "Specificity:\n"
     ]
    }
   ],
   "source": [
    "test_df.loc[test_df['Prob_Score']>=thresholds[np.argmax(precision)],'isFraud_Pred']=1\n",
    "test_df.loc[test_df['Prob_Score']<thresholds[np.argmax(precision)],'isFraud_Pred']=0\n",
    "test_df['isFraud_Pred']=test_df['isFraud_Pred'].astype(int)\n",
    "cm1=confusion_matrix(test_df[target],test_df['isFraud_Pred'])\n",
    "print('Confusion Matrix: \\n',cm1)\n",
    "total=sum(sum(cm1))\n",
    "print(\"\\nAccuracy: {}\" .format(round(((cm1[0,0]+cm1[1,1])/total)*100,2)))\n",
    "print(\"\\nSensitivity/Recall/TPR: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[1,0]))*100,2)))\n",
    "print(\"\\nPrecision: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[0,1]))*100,2)))\n",
    "print(\"\\nSpecificity:\" .format(round((cm1[0,0]/(cm1[0,0]+cm1[0,1]))*100,2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix: \n",
      " [[79582     0]\n",
      " [  200  1848]]\n",
      "\n",
      "Accuracy: 99.75\n",
      "\n",
      "Sensitivity/Recall/TPR: 90.23\n",
      "\n",
      "Precision: 100.0\n",
      "\n",
      "Specificity:\n"
     ]
    }
   ],
   "source": [
    "df_oot_val.loc[df_oot_val['Prob_Score']>=thresholds[np.argmax(precision)],'isFraud_Pred']=1\n",
    "df_oot_val.loc[df_oot_val['Prob_Score']<thresholds[np.argmax(precision)],'isFraud_Pred']=0\n",
    "df_oot_val['isFraud_Pred']=df_oot_val['isFraud_Pred'].astype(int)\n",
    "cm1=confusion_matrix(df_oot_val[target],df_oot_val['isFraud_Pred'])\n",
    "print('Confusion Matrix: \\n',cm1)\n",
    "total=sum(sum(cm1))\n",
    "print(\"\\nAccuracy: {}\" .format(round(((cm1[0,0]+cm1[1,1])/total)*100,2)))\n",
    "print(\"\\nSensitivity/Recall/TPR: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[1,0]))*100,2)))\n",
    "print(\"\\nPrecision: {}\" .format(round((cm1[1,1]/(cm1[1,1]+cm1[0,1]))*100,2)))\n",
    "print(\"\\nSpecificity:\" .format(round((cm1[0,0]/(cm1[0,0]+cm1[0,1]))*100,2)))"
   ]
  },
  {
   "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
}