How to run uplift_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "pd.options.mode.chained_assignment = None\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score, f1_score, average_precision_score\n",
    "from sklearn.metrics import roc_auc_score, roc_curve, auc, make_scorer, classification_report\n",
    "from sklearn.metrics import mean_squared_error, r2_score\n",
    "from sklearn.impute import KNNImputer\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.tree import DecisionTreeRegressor, DecisionTreeClassifier\n",
    "from sklearn.utils import resample\n",
    "from xgboost import XGBClassifier, XGBRegressor, plot_tree\n",
    "from xgboost import plot_importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load in our uplift model library\n",
    "from main import uplift_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#==================================================================#\n",
    "# Here we need to first prepare our data for modelling. Please     #\n",
    "# keep in mind that your dataframe (here referred to as train in   #\n",
    "# our code) must have all features, treatment column and target    #\n",
    "# column. We will then separate them out into X, t and y           #\n",
    "# respectively.                                                    #\n",
    "#                                                                  #\n",
    "# Here I am using a sample dataset from Kaggle which can be found  #\n",
    "# via link: https://www.kaggle.com/davinwijaya/customer-retention  #\n",
    "#==================================================================#\n",
    "\n",
    "# Load csv file\n",
    "df = pd.read_csv(\"Data/data.csv\")\n",
    "df['party_id'] = df.index\n",
    "df = df[df['offer'].isin(['Discount', 'No Offer'])]\n",
    "\n",
    "# Treatment feature\n",
    "df['treatment'] = [1 if x == 'Discount' else 0 for x in df['offer']]\n",
    "\n",
    "# Create a dataframe of features, treatment, target that is ready for training\n",
    "train = pd.DataFrame()\n",
    "train = pd.concat([train, \n",
    "                   df[['party_id', 'recency', 'history', 'used_discount', 'used_bogo',\n",
    "                       'is_referral']],\n",
    "                   pd.get_dummies(df[['zip_code', 'channel']]),\n",
    "                   df[['treatment', 'conversion']]], sort=False, axis=1).reset_index()\n",
    "\n",
    "# Create X, t and y\n",
    "X = train.iloc[:,2:-2]\n",
    "t = train['treatment']\n",
    "y = train['conversion']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Invoke our uplift_model class\n",
    "mdl = uplift_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mu0 XGBoost Classifier for control: [Accuracy: 0.7491, F1-score: 0.7648, Gini: 0.6476]\n",
      "mu1 XGBoost Classifier for treatment: [Accuracy: 0.6678, F1-score: 0.6760, Gini: 0.4744]\n",
      "tau0 XGBoost Regressor: [Mean Squared Error: 0.1028, R2-score: 0.0606]\n",
      "tau1 XGBoost Regressor: [Mean Squared Error: 0.1699, R2-score: 0.0599]\n"
     ]
    }
   ],
   "source": [
    "# Run xlearner_uplift in uplift_model class\n",
    "final, clf0, clf1, reg0, reg1 = mdl.xlearner_uplift(X, t, y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Here just to demonstrate the results we cut our population into deciles\n",
    "final['decile'] = pd.qcut(final['uplift'], 10, labels=False)\n",
    "final['decile'] = 10-final['decile']"
   ]
  },
  {
   "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mu0</th>\n",
       "      <th>mu1</th>\n",
       "      <th>tau0_ite</th>\n",
       "      <th>tau1_ite</th>\n",
       "      <th>uplift</th>\n",
       "      <th>conversion</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "      <th>mean</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>decile</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.261386</td>\n",
       "      <td>0.581696</td>\n",
       "      <td>0.469899</td>\n",
       "      <td>0.024860</td>\n",
       "      <td>0.247380</td>\n",
       "      <td>0.241671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.292670</td>\n",
       "      <td>0.505546</td>\n",
       "      <td>0.412640</td>\n",
       "      <td>-0.100496</td>\n",
       "      <td>0.156072</td>\n",
       "      <td>0.152781</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.303523</td>\n",
       "      <td>0.464505</td>\n",
       "      <td>0.376971</td>\n",
       "      <td>-0.132932</td>\n",
       "      <td>0.122020</td>\n",
       "      <td>0.122506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.310775</td>\n",
       "      <td>0.433536</td>\n",
       "      <td>0.354977</td>\n",
       "      <td>-0.155993</td>\n",
       "      <td>0.099492</td>\n",
       "      <td>0.115520</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.348867</td>\n",
       "      <td>0.428755</td>\n",
       "      <td>0.345081</td>\n",
       "      <td>-0.187002</td>\n",
       "      <td>0.079039</td>\n",
       "      <td>0.112597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.374741</td>\n",
       "      <td>0.417335</td>\n",
       "      <td>0.329273</td>\n",
       "      <td>-0.214400</td>\n",
       "      <td>0.057437</td>\n",
       "      <td>0.106992</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.403830</td>\n",
       "      <td>0.405946</td>\n",
       "      <td>0.310858</td>\n",
       "      <td>-0.242250</td>\n",
       "      <td>0.034304</td>\n",
       "      <td>0.113823</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.452053</td>\n",
       "      <td>0.406307</td>\n",
       "      <td>0.298166</td>\n",
       "      <td>-0.282605</td>\n",
       "      <td>0.007781</td>\n",
       "      <td>0.125323</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.514443</td>\n",
       "      <td>0.412325</td>\n",
       "      <td>0.285462</td>\n",
       "      <td>-0.340256</td>\n",
       "      <td>-0.027397</td>\n",
       "      <td>0.138967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.597930</td>\n",
       "      <td>0.380775</td>\n",
       "      <td>0.177819</td>\n",
       "      <td>-0.400768</td>\n",
       "      <td>-0.111475</td>\n",
       "      <td>0.214403</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             mu0       mu1  tau0_ite  tau1_ite    uplift conversion\n",
       "            mean      mean      mean      mean      mean       mean\n",
       "decile                                                             \n",
       "1       0.261386  0.581696  0.469899  0.024860  0.247380   0.241671\n",
       "2       0.292670  0.505546  0.412640 -0.100496  0.156072   0.152781\n",
       "3       0.303523  0.464505  0.376971 -0.132932  0.122020   0.122506\n",
       "4       0.310775  0.433536  0.354977 -0.155993  0.099492   0.115520\n",
       "5       0.348867  0.428755  0.345081 -0.187002  0.079039   0.112597\n",
       "6       0.374741  0.417335  0.329273 -0.214400  0.057437   0.106992\n",
       "7       0.403830  0.405946  0.310858 -0.242250  0.034304   0.113823\n",
       "8       0.452053  0.406307  0.298166 -0.282605  0.007781   0.125323\n",
       "9       0.514443  0.412325  0.285462 -0.340256 -0.027397   0.138967\n",
       "10      0.597930  0.380775  0.177819 -0.400768 -0.111475   0.214403"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# A summary table to show uplift by decile\n",
    "final[['mu0', 'mu1', 'tau0_ite', 'tau1_ite', 'uplift', 'conversion', 'decile']].groupby(['decile']).agg(['mean'])"
   ]
  }
 ],
 "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": 4
}