05-model-training.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "0eea58ce-de6f-466a-8f55-3b829b1b1702",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Import Libraries\n",
    "import mlflow\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import xgboost as xgb\n",
    "import tensorflow as tf\n",
    "from utils import read_yaml\n",
    "from sklearn.svm import SVC\n",
    "import matplotlib.pyplot as plt\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from mlflow.models import infer_signature\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from mlflow.data.pandas_dataset import PandasDataset\n",
    "from imblearn.under_sampling import OneSidedSelection\n",
    "from sklearn.preprocessing import LabelEncoder, StandardScaler\n",
    "from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier\n",
    "from sklearn.model_selection import train_test_split,cross_val_score,RandomizedSearchCV\n",
    "from sklearn.metrics import (roc_curve, roc_auc_score, \n",
    "                            classification_report, accuracy_score, \n",
    "                            precision_score, recall_score, \n",
    "                            f1_score)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1005b420-86c4-415d-9ed4-3ea45c648b58",
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('data/train1.csv')\n",
    "test = pd.read_csv('data/test1.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "8d2c6137-01c6-499a-8a61-69470f83ef46",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>fnlwgt</th>\n",
       "      <th>education</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>race</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "      <th>income</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.044302</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02174</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.048238</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.122449</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.138113</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.642857</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.151068</td>\n",
       "      <td>0.866667</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.642857</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.221488</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        age  workclass    fnlwgt  education  education_num  marital_status  \\\n",
       "0  0.333333        0.0  0.044302   0.066667       0.800000        1.000000   \n",
       "1  0.333333        0.0  0.048238   0.066667       0.800000        0.666667   \n",
       "2  0.333333        1.0  0.138113   0.000000       0.533333        0.333333   \n",
       "3  0.333333        1.0  0.151068   0.866667       0.400000        0.666667   \n",
       "4  0.000000        1.0  0.221488   0.066667       0.800000        0.666667   \n",
       "\n",
       "   occupation  relationship  race  sex  capital_gain  capital_loss  \\\n",
       "0    0.285714           0.8   1.0  0.0       0.02174           0.0   \n",
       "1    0.071429           0.2   1.0  0.0       0.00000           0.0   \n",
       "2    0.642857           0.8   1.0  0.0       0.00000           0.0   \n",
       "3    0.642857           0.2   0.0  0.0       0.00000           0.0   \n",
       "4    0.142857           0.6   0.0  1.0       0.00000           0.0   \n",
       "\n",
       "   hours_per_week  native_country  income  \n",
       "0        0.397959             1.0     0.0  \n",
       "1        0.122449             1.0     0.0  \n",
       "2        0.397959             1.0     0.0  \n",
       "3        0.397959             1.0     0.0  \n",
       "4        0.397959             0.0     0.0  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ee112344-cf12-4be3-a5a8-b2efaae7146b",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>fnlwgt</th>\n",
       "      <th>education</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>race</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "      <th>income</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.144430</td>\n",
       "      <td>0.866667</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.051677</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.219011</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>0.733333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.099418</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.076881</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.060942</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.295918</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        age  workclass    fnlwgt  education  education_num  marital_status  \\\n",
       "0  0.000000        1.0  0.144430   0.866667       0.400000        1.000000   \n",
       "1  0.333333        1.0  0.051677   0.000000       0.533333        0.666667   \n",
       "2  0.000000        0.0  0.219011   0.400000       0.733333        0.666667   \n",
       "3  0.333333        1.0  0.099418   0.133333       0.600000        0.666667   \n",
       "4  0.000000        0.0  0.060942   0.133333       0.600000        1.000000   \n",
       "\n",
       "   occupation  relationship  race  sex  capital_gain  capital_loss  \\\n",
       "0    0.214286           1.0   0.0  0.0      0.000000           0.0   \n",
       "1    0.285714           0.2   1.0  0.0      0.000000           0.0   \n",
       "2    0.428571           0.2   1.0  0.0      0.000000           0.0   \n",
       "3    0.214286           0.2   0.0  0.0      0.076881           0.0   \n",
       "4    0.714286           1.0   1.0  1.0      0.000000           0.0   \n",
       "\n",
       "   hours_per_week  native_country  income  \n",
       "0        0.397959             1.0     0.0  \n",
       "1        0.500000             1.0     0.0  \n",
       "2        0.397959             1.0     1.0  \n",
       "3        0.397959             1.0     1.0  \n",
       "4        0.295918             1.0     0.0  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6a43dd89-e8ce-4df1-8439-b894ca505cee",
   "metadata": {},
   "outputs": [],
   "source": [
    "#X_train = train.drop('income', axis = 1)\n",
    "X_train = train.drop(['income'], axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "53c485ee-f06b-4e71-9c33-30e18fb63042",
   "metadata": {},
   "outputs": [],
   "source": [
    "#X_test = test.drop('income', axis = 1)\n",
    "X_test = test.drop(['income'], axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "465b1014-0c03-4713-a625-6cf91968cab0",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = train['income']\n",
    "y_test = test['income']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "71b52fe7-9cc8-4bdf-b9e4-b38be16ad998",
   "metadata": {},
   "outputs": [],
   "source": [
    "#undersampling the train set\n",
    "under = OneSidedSelection()\n",
    "X_train_usample, y_train_usample = under.fit_resample(X_train, train['income'])\n",
    "\n",
    "\n",
    "#oversampling the train set\n",
    "sm = SMOTE()\n",
    "X_train_osample, y_train_osample = sm.fit_resample(X_train_usample, y_train_usample)\n",
    "\n",
    "X_train = pd.DataFrame(X_train_osample, columns=train.drop(['income'], axis=1).columns)\n",
    "\n",
    "\n",
    "\n",
    "#creating the final train \n",
    "X_train = pd.concat([X_train, y_train_osample], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b0fa5800-a918-4822-a084-7b8c72dbd059",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>fnlwgt</th>\n",
       "      <th>education</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>race</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "      <th>income</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.057742</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.479592</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.044302</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02174</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.048238</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.122449</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.138113</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.642857</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.221488</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        age  workclass    fnlwgt  education  education_num  marital_status  \\\n",
       "0  0.000000        1.0  0.057742   0.066667       0.800000        1.000000   \n",
       "1  0.333333        0.0  0.044302   0.066667       0.800000        1.000000   \n",
       "2  0.333333        0.0  0.048238   0.066667       0.800000        0.666667   \n",
       "3  0.333333        1.0  0.138113   0.000000       0.533333        0.333333   \n",
       "4  0.000000        1.0  0.221488   0.066667       0.800000        0.666667   \n",
       "\n",
       "   occupation  relationship  race  sex  capital_gain  capital_loss  \\\n",
       "0    0.285714           0.8   1.0  1.0       0.00000           0.0   \n",
       "1    0.285714           0.8   1.0  0.0       0.02174           0.0   \n",
       "2    0.071429           0.2   1.0  0.0       0.00000           0.0   \n",
       "3    0.642857           0.8   1.0  0.0       0.00000           0.0   \n",
       "4    0.142857           0.6   0.0  1.0       0.00000           0.0   \n",
       "\n",
       "   hours_per_week  native_country  income  \n",
       "0        0.479592             1.0     0.0  \n",
       "1        0.397959             1.0     0.0  \n",
       "2        0.122449             1.0     0.0  \n",
       "3        0.397959             1.0     0.0  \n",
       "4        0.397959             0.0     0.0  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "8aedc68b-a064-4e35-aa96-ed908bae89da",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 45780 entries, 0 to 45779\n",
      "Data columns (total 15 columns):\n",
      " #   Column          Non-Null Count  Dtype  \n",
      "---  ------          --------------  -----  \n",
      " 0   age             45780 non-null  float64\n",
      " 1   workclass       45780 non-null  float64\n",
      " 2   fnlwgt          45780 non-null  float64\n",
      " 3   education       45780 non-null  float64\n",
      " 4   education_num   45780 non-null  float64\n",
      " 5   marital_status  45780 non-null  float64\n",
      " 6   occupation      45780 non-null  float64\n",
      " 7   relationship    45780 non-null  float64\n",
      " 8   race            45780 non-null  float64\n",
      " 9   sex             45780 non-null  float64\n",
      " 10  capital_gain    45780 non-null  float64\n",
      " 11  capital_loss    45780 non-null  float64\n",
      " 12  hours_per_week  45780 non-null  float64\n",
      " 13  native_country  45780 non-null  float64\n",
      " 14  income          45780 non-null  float64\n",
      "dtypes: float64(15)\n",
      "memory usage: 5.2 MB\n"
     ]
    }
   ],
   "source": [
    "X_train.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a2513eb3-51c6-4ec9-835e-6bc8e42cd0c6",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = X_train['income']\n",
    "X_train.drop(['income', 'fnlwgt', 'race', 'education'], inplace=True, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "180a0753-0663-4b67-99c2-6935fb6d0d0c",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test.drop(['fnlwgt', 'race', 'education'], inplace=True, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9b2e9cf3-1b5f-4985-b2d4-814fc94be756",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.479592</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.021740</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.200000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.122449</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.642857</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45775</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.866667</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.200000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.150242</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.662818</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45776</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.128443</td>\n",
       "      <td>0.200000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.436639</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45777</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.200000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.026547</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45778</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.548154</td>\n",
       "      <td>0.407438</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.755537</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500196</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45779</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.200000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.075977</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>45780 rows × 11 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            age  workclass  education_num  marital_status  occupation  \\\n",
       "0      0.000000        1.0       0.800000        1.000000    0.285714   \n",
       "1      0.333333        0.0       0.800000        1.000000    0.285714   \n",
       "2      0.333333        0.0       0.800000        0.666667    0.071429   \n",
       "3      0.333333        1.0       0.533333        0.333333    0.642857   \n",
       "4      0.000000        1.0       0.800000        0.666667    0.142857   \n",
       "...         ...        ...            ...             ...         ...   \n",
       "45775  0.333333        1.0       0.866667        0.666667    0.071429   \n",
       "45776  0.333333        1.0       0.800000        0.666667    0.128443   \n",
       "45777  0.333333        0.0       1.000000        0.666667    0.142857   \n",
       "45778  0.333333        1.0       0.548154        0.407438    0.071429   \n",
       "45779  0.000000        0.0       0.533333        0.666667    1.000000   \n",
       "\n",
       "       relationship  sex  capital_gain  capital_loss  hours_per_week  \\\n",
       "0          0.800000  1.0      0.000000      0.000000        0.479592   \n",
       "1          0.800000  0.0      0.021740      0.000000        0.397959   \n",
       "2          0.200000  0.0      0.000000      0.000000        0.122449   \n",
       "3          0.800000  0.0      0.000000      0.000000        0.397959   \n",
       "4          0.600000  1.0      0.000000      0.000000        0.397959   \n",
       "...             ...  ...           ...           ...             ...   \n",
       "45775      0.200000  0.0      0.150242      0.000000        0.662818   \n",
       "45776      0.200000  0.0      0.000000      0.436639        0.500000   \n",
       "45777      0.200000  0.0      0.026547      0.000000        0.500000   \n",
       "45778      0.755537  1.0      0.000000      0.500196        0.397959   \n",
       "45779      0.200000  0.0      0.075977      0.000000        0.397959   \n",
       "\n",
       "       native_country  \n",
       "0                 1.0  \n",
       "1                 1.0  \n",
       "2                 1.0  \n",
       "3                 1.0  \n",
       "4                 0.0  \n",
       "...               ...  \n",
       "45775             1.0  \n",
       "45776             1.0  \n",
       "45777             1.0  \n",
       "45778             1.0  \n",
       "45779             1.0  \n",
       "\n",
       "[45780 rows x 11 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c29534dc-4e4e-4809-9aff-3702bdd85a54",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.733333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.076881</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.295918</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        age  workclass  education_num  marital_status  occupation  \\\n",
       "0  0.000000        1.0       0.400000        1.000000    0.214286   \n",
       "1  0.333333        1.0       0.533333        0.666667    0.285714   \n",
       "2  0.000000        0.0       0.733333        0.666667    0.428571   \n",
       "3  0.333333        1.0       0.600000        0.666667    0.214286   \n",
       "4  0.000000        0.0       0.600000        1.000000    0.714286   \n",
       "\n",
       "   relationship  sex  capital_gain  capital_loss  hours_per_week  \\\n",
       "0           1.0  0.0      0.000000           0.0        0.397959   \n",
       "1           0.2  0.0      0.000000           0.0        0.500000   \n",
       "2           0.2  0.0      0.000000           0.0        0.397959   \n",
       "3           0.2  0.0      0.076881           0.0        0.397959   \n",
       "4           1.0  1.0      0.000000           0.0        0.295918   \n",
       "\n",
       "   native_country  \n",
       "0             1.0  \n",
       "1             1.0  \n",
       "2             1.0  \n",
       "3             1.0  \n",
       "4             1.0  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "26ddd876-442f-4b52-b6e1-155701005365",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    22890\n",
       "1.0    22890\n",
       "Name: income, dtype: int64"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "98eeec94-699f-4425-9eb3-e88a42ff5423",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "yaml file: params.yaml loaded successfully\n"
     ]
    }
   ],
   "source": [
    "configbox = read_yaml('params.yaml')\n",
    "xgb_params = configbox.XGBOOST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "48508595-db36-4a49-88e2-bdd8c7273fb4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ConfigBox({'n_estimators': 128, 'learning_rate': 0.1, 'max_depth': 6})"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb_params"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f5d33c9-3515-4c2b-ab79-a9a7f452c0df",
   "metadata": {},
   "source": [
    "### XGB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "fbd58a0f-eafb-4361-85a7-94638ddca756",
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_classifier = xgb.XGBClassifier(n_estimators = xgb_params.n_estimators,\n",
    "                                   learning_rate = xgb_params.learning_rate,\n",
    "                                  max_depth = xgb_params.max_depth)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "399d438a-f41d-438d-be96-015abf37e877",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8968233074459216\n"
     ]
    }
   ],
   "source": [
    "xgb_classifier.fit(X_train, y_train)\n",
    "y_pred = xgb_classifier.predict(X_test)\n",
    "y_pred_train = xgb_classifier.predict(X_train)\n",
    "\n",
    "accuracy_train = accuracy_score(y_train, y_pred_train)\n",
    "print(accuracy_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "a96a7d77-0e51-4e35-aacc-c30e4f21f308",
   "metadata": {},
   "outputs": [],
   "source": [
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f_score = f1_score(y_test, y_pred)\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "auc = roc_auc_score(y_test,y_pred)\n",
    "\n",
    "metrics = {'Precision': precision, 'Recall': recall, 'f1_score': f_score, 'AUC': auc, 'accuracy': accuracy}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "498d8e82-8b2d-4343-8acb-aa50449dd065",
   "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>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0.0</th>\n",
       "      <td>0.928197</td>\n",
       "      <td>0.826458</td>\n",
       "      <td>0.874378</td>\n",
       "      <td>12435.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <td>0.585717</td>\n",
       "      <td>0.793292</td>\n",
       "      <td>0.673882</td>\n",
       "      <td>3846.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>accuracy</th>\n",
       "      <td>0.818623</td>\n",
       "      <td>0.818623</td>\n",
       "      <td>0.818623</td>\n",
       "      <td>0.818623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.756957</td>\n",
       "      <td>0.809875</td>\n",
       "      <td>0.774130</td>\n",
       "      <td>16281.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.847294</td>\n",
       "      <td>0.818623</td>\n",
       "      <td>0.827015</td>\n",
       "      <td>16281.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              precision    recall  f1-score       support\n",
       "0.0            0.928197  0.826458  0.874378  12435.000000\n",
       "1.0            0.585717  0.793292  0.673882   3846.000000\n",
       "accuracy       0.818623  0.818623  0.818623      0.818623\n",
       "macro avg      0.756957  0.809875  0.774130  16281.000000\n",
       "weighted avg   0.847294  0.818623  0.827015  16281.000000"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "report = classification_report(y_test, y_pred, output_dict=True)\n",
    "df = pd.DataFrame(report).transpose()\n",
    "df "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "222775f6-22ab-4f3a-b364-4ca71a5403a2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.92      0.87      0.89     22791\n",
      "         1.0       0.88      0.92      0.90     22791\n",
      "\n",
      "    accuracy                           0.90     45582\n",
      "   macro avg       0.90      0.90      0.90     45582\n",
      "weighted avg       0.90      0.90      0.90     45582\n",
      "\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.93      0.83      0.87     12435\n",
      "         1.0       0.59      0.79      0.67      3846\n",
      "\n",
      "    accuracy                           0.82     16281\n",
      "   macro avg       0.76      0.81      0.77     16281\n",
      "weighted avg       0.85      0.82      0.83     16281\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_train,y_pred_train))\n",
    "print(classification_report(y_test,y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "919f8f2a-8fca-4855-8f4b-253570789f9a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8098746555411072\n",
      "0.5\n"
     ]
    }
   ],
   "source": [
    "xgboost_auc=roc_auc_score(y_test,y_pred)\n",
    "r_probs = [0 for _ in range(len(y_test))]\n",
    "random_auc = roc_auc_score(y_test, r_probs)\n",
    "print(xgboost_auc)\n",
    "print(random_auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "b18079e6-6078-4d3c-914d-af5f4e2c812a",
   "metadata": {},
   "outputs": [],
   "source": [
    "probs = xgb_classifier.predict_proba(X_test)\n",
    "# Keep probabilities for one class\n",
    "probs = probs[:,1]\n",
    "\n",
    "# Get false and True positive rate\n",
    "r_fpr, r_tpr, _ = roc_curve(y_test,probs)\n",
    "x = np.arange(0,1.01,0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "d6a44940-8fdb-4aec-8bf2-a2f3a4d702dd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(r_fpr, r_tpr, label='XGBoost (AUROC = %0.3f)' % xgboost_auc)\n",
    "plt.plot(x,x, linestyle='--', label='Random prediction (AUROC = %0.3f)' % random_auc)\n",
    "plt.legend()\n",
    "plt.title('ROC CURVE')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.savefig(f'visualization/roc_curve{xgboost_auc}.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "ee67be92-8b7f-4d40-9e0e-c67f2f9bd960",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Windows 10\\.conda\\envs\\tensorflow_env\\lib\\site-packages\\mlflow\\data\\dataset_source_registry.py:150: UserWarning: The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource. MLflow will assume that this is a LocalArtifactDatasetSource source.\n",
      "  return _dataset_source_registry.resolve(\n",
      "C:\\Users\\Windows 10\\.conda\\envs\\tensorflow_env\\lib\\site-packages\\mlflow\\data\\dataset_source_registry.py:150: UserWarning: The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource. MLflow will assume that this is a LocalArtifactDatasetSource source.\n",
      "  return _dataset_source_registry.resolve(\n"
     ]
    }
   ],
   "source": [
    "# Construct an MLflow PandasDataset from the Pandas DataFrame, and specify the source\n",
    "train_dataset: PandasDataset = mlflow.data.from_pandas(train, source= 'data/train1.csv')\n",
    "# test\n",
    "test_dataset: PandasDataset = mlflow.data.from_pandas(test, source= 'data/test1.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "112618b9-89ac-4af7-a341-96a255ae2481",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Windows 10\\.conda\\envs\\tensorflow_env\\lib\\site-packages\\xgboost\\core.py:160: UserWarning: [13:06:14] WARNING: C:\\buildkite-agent\\builds\\buildkite-windows-cpu-autoscaling-group-i-0b3782d1791676daf-1\\xgboost\\xgboost-ci-windows\\src\\c_api\\c_api.cc:1240: Saving into deprecated binary model format, please consider using `json` or `ubj`. Model format will default to JSON in XGBoost 2.2 if not specified.\n",
      "  warnings.warn(smsg, UserWarning)\n",
      "C:\\Users\\Windows 10\\.conda\\envs\\tensorflow_env\\lib\\site-packages\\_distutils_hack\\__init__.py:33: UserWarning: Setuptools is replacing distutils.\n",
      "  warnings.warn(\"Setuptools is replacing distutils.\")\n",
      "Registered model 'xgboost' already exists. Creating a new version of this model...\n",
      "2024/03/21 13:06:22 INFO mlflow.store.model_registry.abstract_store: Waiting up to 300 seconds for model version to finish creation. Model name: xgboost, version 10\n",
      "Created version '10' of model 'xgboost'.\n"
     ]
    }
   ],
   "source": [
    "\n",
    "mlflow.set_tracking_uri('http://localhost:5000')\n",
    "\n",
    "mlflow.set_experiment(\"XGBOOST EXPERIMENTS\")\n",
    "\n",
    "with mlflow.start_run() as run:\n",
    "    mlflow.log_params(xgb_params)\n",
    "    mlflow.log_metrics(metrics)\n",
    "    \n",
    "    # Set a tag that we can use to remind ourselves what this run was for\n",
    "    mlflow.set_tag(\"Some Columns\", \"no race, education, fnlwgt undersampling and over sampling\")\n",
    "    \n",
    "    # Log the dataset to the MLflow Run. Specify the \"training\" context to indicate that the\n",
    "    # dataset is used for model training\n",
    "    mlflow.log_input(train_dataset, context=\"training\")\n",
    "    mlflow.log_input(test_dataset, context = \"testing\")\n",
    "    mlflow.log_artifact(f'visualization/roc_curve{xgboost_auc}.png')\n",
    "\n",
    "    \n",
    "    # Infer the model signature\n",
    "    signature = infer_signature(X_train, y_pred_train)\n",
    "    \n",
    "    # Log the model\n",
    "    model_info = mlflow.xgboost.log_model(\n",
    "        xgb_model= xgb_classifier,\n",
    "        artifact_path = \"income_class\",\n",
    "        signature = signature,\n",
    "        input_example = X_train,\n",
    "        registered_model_name = \"xgboost\",\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "0e2eefad-d054-466a-9a31-cf267bfc26b3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Inference successful!\n",
      "{'predictions': [0, 1, 0, 1, 0, 0, 0, 1, 0, 0]}\n"
     ]
    }
   ],
   "source": [
    "import requests\n",
    "import pandas as pd \n",
    "\n",
    "# Assuming X_test is a pandas DataFrame\n",
    "lst = X_test[:10].to_dict(orient=\"records\")  # Convert DataFrame to a list of dictionaries\n",
    "inference_request = {\n",
    "    \"inputs\": lst\n",
    "}\n",
    "endpoint = \"http://localhost:1234/invocations\"\n",
    "response = requests.post(endpoint, json=inference_request)\n",
    "\n",
    "if response.status_code == 200:\n",
    "    print(\"Inference successful!\")\n",
    "    print(response.json())\n",
    "else:\n",
    "    print(f\"Error: {response.status_code} - {response.text}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "9cd66886-a717-4985-a824-9c90543f593f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.0\n",
       "1    0.0\n",
       "2    1.0\n",
       "3    1.0\n",
       "4    0.0\n",
       "5    0.0\n",
       "6    0.0\n",
       "7    1.0\n",
       "8    0.0\n",
       "9    0.0\n",
       "Name: income, dtype: float64"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "734fd214-1292-4270-9593-9ecf1c4a72ce",
   "metadata": {},
   "source": [
    "## DNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "beb1493e-160c-4017-9d9d-37b3ff399f74",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_train_examples, input_dim = X_train.shape\n",
    "input_dim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "91d76baf-a409-44a9-a709-174f1514423e",
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 128\n",
    "num_train_examples, input_dim = X_train.shape\n",
    "# Number of times to pass through data\n",
    "num_epcohs = 20\n",
    "\n",
    "input = tf.keras.Input(shape=(11,))\n",
    "#x = tf.keras.layers.concatenate(input)\n",
    "x = tf.keras.layers.Dense(100, activation='relu', kernel_initializer= 'uniform')(input)\n",
    "x = tf.keras.layers.Dense(75, activation='relu')(x)\n",
    "x = tf.keras.layers.Dense(50, activation='relu')(x)\n",
    "x = tf.keras.layers.Dense(25, activation='relu')(x)\n",
    "output = tf.keras.layers.Dense(1, activation='sigmoid')(x)\n",
    "\n",
    "model = tf.keras.Model(inputs = input, outputs= output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "f7a12155-7821-4022-9aaf-8a9c2c8cebaf",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:absl:`lr` is deprecated in Keras optimizer, please use `learning_rate` or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.RMSprop.\n"
     ]
    }
   ],
   "source": [
    "learning_rate = .01\n",
    "\n",
    "# Optimizer\n",
    "optimizer = tf.keras.optimizers.RMSprop(lr=learning_rate)\n",
    "# Compile Keras model\n",
    "model.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "cd0ef06b-8d18-4629-86b9-4e9897cc7fd5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Setup Learning Rate decay.\n",
    "lr_decay_cb = tf.keras.callbacks.LearningRateScheduler(\n",
    "    lambda epoch: learning_rate + 0.02 * (0.5 ** (1 + epoch)),\n",
    "    verbose=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "8de21165-a006-45f8-998d-6b4ae19ac3cc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 1: LearningRateScheduler setting learning rate to 0.02.\n",
      "Epoch 1/20\n",
      "357/357 [==============================] - 2s 4ms/step - loss: 0.3942 - accuracy: 0.8198 - lr: 0.0200\n",
      "\n",
      "Epoch 2: LearningRateScheduler setting learning rate to 0.015.\n",
      "Epoch 2/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3713 - accuracy: 0.8320 - lr: 0.0150\n",
      "\n",
      "Epoch 3: LearningRateScheduler setting learning rate to 0.0125.\n",
      "Epoch 3/20\n",
      "357/357 [==============================] - 1s 3ms/step - loss: 0.3628 - accuracy: 0.8369 - lr: 0.0125\n",
      "\n",
      "Epoch 4: LearningRateScheduler setting learning rate to 0.01125.\n",
      "Epoch 4/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3595 - accuracy: 0.8382 - lr: 0.0113\n",
      "\n",
      "Epoch 5: LearningRateScheduler setting learning rate to 0.010625.\n",
      "Epoch 5/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3567 - accuracy: 0.8401 - lr: 0.0106\n",
      "\n",
      "Epoch 6: LearningRateScheduler setting learning rate to 0.0103125.\n",
      "Epoch 6/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3544 - accuracy: 0.8409 - lr: 0.0103\n",
      "\n",
      "Epoch 7: LearningRateScheduler setting learning rate to 0.01015625.\n",
      "Epoch 7/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3522 - accuracy: 0.8429 - lr: 0.0102\n",
      "\n",
      "Epoch 8: LearningRateScheduler setting learning rate to 0.010078125.\n",
      "Epoch 8/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3511 - accuracy: 0.8442 - lr: 0.0101\n",
      "\n",
      "Epoch 9: LearningRateScheduler setting learning rate to 0.0100390625.\n",
      "Epoch 9/20\n",
      "357/357 [==============================] - 2s 4ms/step - loss: 0.3492 - accuracy: 0.8457 - lr: 0.0100\n",
      "\n",
      "Epoch 10: LearningRateScheduler setting learning rate to 0.01001953125.\n",
      "Epoch 10/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3483 - accuracy: 0.8441 - lr: 0.0100\n",
      "\n",
      "Epoch 11: LearningRateScheduler setting learning rate to 0.010009765625.\n",
      "Epoch 11/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3480 - accuracy: 0.8448 - lr: 0.0100\n",
      "\n",
      "Epoch 12: LearningRateScheduler setting learning rate to 0.010004882812500001.\n",
      "Epoch 12/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3467 - accuracy: 0.8458 - lr: 0.0100\n",
      "\n",
      "Epoch 13: LearningRateScheduler setting learning rate to 0.01000244140625.\n",
      "Epoch 13/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3460 - accuracy: 0.8457 - lr: 0.0100\n",
      "\n",
      "Epoch 14: LearningRateScheduler setting learning rate to 0.010001220703125.\n",
      "Epoch 14/20\n",
      "357/357 [==============================] - 2s 4ms/step - loss: 0.3449 - accuracy: 0.8459 - lr: 0.0100\n",
      "\n",
      "Epoch 15: LearningRateScheduler setting learning rate to 0.0100006103515625.\n",
      "Epoch 15/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3446 - accuracy: 0.8461 - lr: 0.0100\n",
      "\n",
      "Epoch 16: LearningRateScheduler setting learning rate to 0.01000030517578125.\n",
      "Epoch 16/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3444 - accuracy: 0.8460 - lr: 0.0100\n",
      "\n",
      "Epoch 17: LearningRateScheduler setting learning rate to 0.010000152587890625.\n",
      "Epoch 17/20\n",
      "357/357 [==============================] - 2s 5ms/step - loss: 0.3428 - accuracy: 0.8471 - lr: 0.0100\n",
      "\n",
      "Epoch 18: LearningRateScheduler setting learning rate to 0.010000076293945313.\n",
      "Epoch 18/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3445 - accuracy: 0.8466 - lr: 0.0100\n",
      "\n",
      "Epoch 19: LearningRateScheduler setting learning rate to 0.010000038146972657.\n",
      "Epoch 19/20\n",
      "357/357 [==============================] - 1s 4ms/step - loss: 0.3431 - accuracy: 0.8458 - lr: 0.0100\n",
      "\n",
      "Epoch 20: LearningRateScheduler setting learning rate to 0.010000019073486329.\n",
      "Epoch 20/20\n",
      "357/357 [==============================] - 2s 4ms/step - loss: 0.3424 - accuracy: 0.8478 - lr: 0.0100\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x2d1ab3488e0>"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(\n",
    "    X_train,\n",
    "    y_train,\n",
    "    epochs = num_epcohs,\n",
    "    batch_size= 128,\n",
    "    steps_per_epoch=int(num_train_examples / batch_size),\n",
    "    callbacks = [lr_decay_cb]  \n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "c2d8b335-77a4-446a-873f-b33dbc1d2d85",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "509/509 [==============================] - 1s 3ms/step\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "96925582-463a-43f8-9ff0-da1a28baece1",
   "metadata": {},
   "outputs": [],
   "source": [
    "classes = y_pred > 0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "id": "c734f4d5-c4a2-4562-9a9c-f7263ebae2ab",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7873594988022848"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(y_test, classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "68f6fc3b-cdad-41ba-aca2-e851a79d7115",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8132042000618505"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_test, classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "0b990bd8-421e-4899-9ef4-58bd826d33bf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.657023974638399"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_test, classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "7d7ae9e8-104a-4b45-b67f-ca227c0b71d5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.5\n"
     ]
    }
   ],
   "source": [
    "r_probs = [0 for _ in range(len(y_test))]\n",
    "random_auc = roc_auc_score(y_test, r_probs)\n",
    "#print(xgauc)\n",
    "print(random_auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "858200e4-3887-4dad-a043-a5136c74badf",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set tracking uri\n",
    "mlflow.set_tracking_uri('http://localhost:5000')\n",
    "\n",
    "# Get the model uri you need\n",
    "model_uri = 'runs:/97c3446b013742c6988f0ab7eeb5f304/income_class'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ccc2f8e6-666c-49fc-8071-710c03ab533b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the model using mlflow.xgboost.load_model\n",
    "loaded_model = mlflow.xgboost.load_model(model_uri)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "49f40ad6-96ba-45f5-b3b2-f16c69af002d",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# While not directly recommended for deployment due to limitations,\n",
    "# you can use pickle for temporary use or exploration (avoid for production)\n",
    "# But I will be using it here\n",
    "import pickle\n",
    "with open(\"income-pred/app/income.pkl\", \"wb\") as f:\n",
    "    pickle.dump(loaded_model, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "10af2e57-2e84-444c-b974-882fc54f57b0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "5459d7ba-9bbd-4292-a9d8-9f17d347a380",
   "metadata": {},
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "24324037-7c6f-4ab6-bf76-49766fd0034d",
   "metadata": {},
   "outputs": [],
   "source": [
    "rf=RandomForestClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8530e1ba-e469-4395-a4b3-2bf11e3d18b7",
   "metadata": {},
   "outputs": [],
   "source": [
    "rf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "56a90761-2dc3-4882-bcf5-60b364b98d3c",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = rf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28b2929a-6386-4f6e-8843-8a7f97ebf9fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "print(classification_report(y_test,y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6eaa021a-c0c8-44cd-942a-da0287a78ad0",
   "metadata": {},
   "outputs": [],
   "source": [
    "import mlflow\n",
    "from mlflow.models import infer_signature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c7c80703-8934-48c7-a576-caf4eab8ad1d",
   "metadata": {},
   "outputs": [],
   "source": [
    "mlflow.set_tracking_uri(uri=\"http://127.0.0.1:8080\")\n",
    "\n",
    "mlflow.set_experiment(\"XGBOOST EXPERIMENTS\")\n",
    "\n",
    "with mlflow.start_run() as run:\n",
    "    mlflow.log_params(rf_params)\n",
    "    mlflow.log_metrics(rf_metrics)\n",
    "    \n",
    "    # Set a tag that we can use to remind ourselves what this run was for\n",
    "    mlflow.set_tag(\"\")\n",
    "    \n",
    "    # Infer the model signature\n",
    "    signature = infer_signature(X_train, y_pred_train)\n",
    "    \n",
    "    # Log the model\n",
    "    model_info = mlflow.xgboost.log_model(\n",
    "        xgb_model= xgb_classifier,\n",
    "        artifact_path = \"income_class\",\n",
    "        signature = signature,\n",
    "        input_example = X_train,\n",
    "        registered_model_name = \"random_forest\",\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3969b576-805c-4eec-adc9-0e232a8d21d4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import mlflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "13f564c1-f8bf-466c-a15d-6f50021852cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "mlflow.set_tracking_uri(uri=\"http://127.0.0.1:8080\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "12ef49dd-701b-4a72-bd7c-ce7deb01fa81",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test = test.drop(\"income\", axis = 1)\n",
    "y_test = test[\"income\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "fbea800d-4c74-4d38-8235-ded6b0e2bc3c",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>fnlwgt</th>\n",
       "      <th>education</th>\n",
       "      <th>education_num</th>\n",
       "      <th>marital_status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>race</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital_gain</th>\n",
       "      <th>capital_loss</th>\n",
       "      <th>hours_per_week</th>\n",
       "      <th>native_country</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.144430</td>\n",
       "      <td>0.866667</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.051677</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.219011</td>\n",
       "      <td>0.400000</td>\n",
       "      <td>0.733333</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.099418</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.214286</td>\n",
       "      <td>0.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.076881</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.060942</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>0.600000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.295918</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16276</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.136723</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.357143</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16277</th>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.208484</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.533333</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>0.4</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16278</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.244762</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.142857</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16279</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.047666</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.357143</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.054551</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.397959</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16280</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.114195</td>\n",
       "      <td>0.066667</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.2</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.602041</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>16281 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            age  workclass    fnlwgt  education  education_num  \\\n",
       "0      0.000000        1.0  0.144430   0.866667       0.400000   \n",
       "1      0.333333        1.0  0.051677   0.000000       0.533333   \n",
       "2      0.000000        0.0  0.219011   0.400000       0.733333   \n",
       "3      0.333333        1.0  0.099418   0.133333       0.600000   \n",
       "4      0.000000        0.0  0.060942   0.133333       0.600000   \n",
       "...         ...        ...       ...        ...            ...   \n",
       "16276  0.333333        1.0  0.136723   0.066667       0.800000   \n",
       "16277  0.666667        0.0  0.208484   0.000000       0.533333   \n",
       "16278  0.333333        1.0  0.244762   0.066667       0.800000   \n",
       "16279  0.333333        1.0  0.047666   0.066667       0.800000   \n",
       "16280  0.000000        0.0  0.114195   0.066667       0.800000   \n",
       "\n",
       "       marital_status  occupation  relationship  race  sex  capital_gain  \\\n",
       "0            1.000000    0.214286           1.0   0.0  0.0      0.000000   \n",
       "1            0.666667    0.285714           0.2   1.0  0.0      0.000000   \n",
       "2            0.666667    0.428571           0.2   1.0  0.0      0.000000   \n",
       "3            0.666667    0.214286           0.2   0.0  0.0      0.076881   \n",
       "4            1.000000    0.714286           1.0   1.0  1.0      0.000000   \n",
       "...               ...         ...           ...   ...  ...           ...   \n",
       "16276        0.333333    0.142857           0.8   1.0  1.0      0.000000   \n",
       "16277        0.000000    0.714286           0.4   0.0  0.0      0.000000   \n",
       "16278        0.666667    0.142857           0.2   1.0  0.0      0.000000   \n",
       "16279        0.333333    0.357143           1.0   0.0  0.0      0.054551   \n",
       "16280        0.666667    0.071429           0.2   1.0  0.0      0.000000   \n",
       "\n",
       "       capital_loss  hours_per_week  native_country  \n",
       "0               0.0        0.397959             1.0  \n",
       "1               0.0        0.500000             1.0  \n",
       "2               0.0        0.397959             1.0  \n",
       "3               0.0        0.397959             1.0  \n",
       "4               0.0        0.295918             1.0  \n",
       "...             ...             ...             ...  \n",
       "16276           0.0        0.357143             1.0  \n",
       "16277           0.0        0.397959             1.0  \n",
       "16278           0.0        0.500000             1.0  \n",
       "16279           0.0        0.397959             1.0  \n",
       "16280           0.0        0.602041             1.0  \n",
       "\n",
       "[16281 rows x 14 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "2d1289e6-250a-4f36-876c-851a5c49f99c",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_data = (X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "b71d2719-edca-45d1-9554-a7491dc5adec",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "with open(\"income-pred/app/data/test_data.pkl\", \"wb\") as f:\n",
    "    pickle.dump(test_data, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "d83a0d42-8fc6-41d2-8e6d-6068a6e08d7c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "X_test = X_test.iloc[6].to_dict()\n",
    "\n",
    "with open(\"income-pred/income-examples/income_example.json\", \"wb\") as f:\n",
    "    json.dumps(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ffc2548d-2d0f-4ad3-b4a8-1a37df62789b",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}