SujanNotebook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.preprocessing import LabelEncoder, OneHotEncoder\n",
    "from datetime import datetime\n",
    "from tqdm import tqdm\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>id</th>\n",
       "      <th>member_id</th>\n",
       "      <th>loan_amnt</th>\n",
       "      <th>funded_amnt</th>\n",
       "      <th>funded_amnt_inv</th>\n",
       "      <th>term</th>\n",
       "      <th>int_rate</th>\n",
       "      <th>installment</th>\n",
       "      <th>grade</th>\n",
       "      <th>sub_grade</th>\n",
       "      <th>...</th>\n",
       "      <th>total_bal_il</th>\n",
       "      <th>il_util</th>\n",
       "      <th>open_rv_12m</th>\n",
       "      <th>open_rv_24m</th>\n",
       "      <th>max_bal_bc</th>\n",
       "      <th>all_util</th>\n",
       "      <th>total_rev_hi_lim</th>\n",
       "      <th>inq_fi</th>\n",
       "      <th>total_cu_tl</th>\n",
       "      <th>inq_last_12m</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1077501</td>\n",
       "      <td>1296599</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>5000.0</td>\n",
       "      <td>4975.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>10.65</td>\n",
       "      <td>162.87</td>\n",
       "      <td>B</td>\n",
       "      <td>B2</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1077430</td>\n",
       "      <td>1314167</td>\n",
       "      <td>2500.0</td>\n",
       "      <td>2500.0</td>\n",
       "      <td>2500.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>15.27</td>\n",
       "      <td>59.83</td>\n",
       "      <td>C</td>\n",
       "      <td>C4</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1077175</td>\n",
       "      <td>1313524</td>\n",
       "      <td>2400.0</td>\n",
       "      <td>2400.0</td>\n",
       "      <td>2400.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>15.96</td>\n",
       "      <td>84.33</td>\n",
       "      <td>C</td>\n",
       "      <td>C5</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1076863</td>\n",
       "      <td>1277178</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>13.49</td>\n",
       "      <td>339.31</td>\n",
       "      <td>C</td>\n",
       "      <td>C1</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1075358</td>\n",
       "      <td>1311748</td>\n",
       "      <td>3000.0</td>\n",
       "      <td>3000.0</td>\n",
       "      <td>3000.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>12.69</td>\n",
       "      <td>67.79</td>\n",
       "      <td>B</td>\n",
       "      <td>B5</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 74 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        id  member_id  loan_amnt  funded_amnt  funded_amnt_inv        term  \\\n",
       "0  1077501    1296599     5000.0       5000.0           4975.0   36 months   \n",
       "1  1077430    1314167     2500.0       2500.0           2500.0   60 months   \n",
       "2  1077175    1313524     2400.0       2400.0           2400.0   36 months   \n",
       "3  1076863    1277178    10000.0      10000.0          10000.0   36 months   \n",
       "4  1075358    1311748     3000.0       3000.0           3000.0   60 months   \n",
       "\n",
       "   int_rate  installment grade sub_grade  ... total_bal_il il_util  \\\n",
       "0     10.65       162.87     B        B2  ...          NaN     NaN   \n",
       "1     15.27        59.83     C        C4  ...          NaN     NaN   \n",
       "2     15.96        84.33     C        C5  ...          NaN     NaN   \n",
       "3     13.49       339.31     C        C1  ...          NaN     NaN   \n",
       "4     12.69        67.79     B        B5  ...          NaN     NaN   \n",
       "\n",
       "  open_rv_12m  open_rv_24m max_bal_bc all_util total_rev_hi_lim inq_fi  \\\n",
       "0         NaN          NaN        NaN      NaN              NaN    NaN   \n",
       "1         NaN          NaN        NaN      NaN              NaN    NaN   \n",
       "2         NaN          NaN        NaN      NaN              NaN    NaN   \n",
       "3         NaN          NaN        NaN      NaN              NaN    NaN   \n",
       "4         NaN          NaN        NaN      NaN              NaN    NaN   \n",
       "\n",
       "  total_cu_tl inq_last_12m  \n",
       "0         NaN          NaN  \n",
       "1         NaN          NaN  \n",
       "2         NaN          NaN  \n",
       "3         NaN          NaN  \n",
       "4         NaN          NaN  \n",
       "\n",
       "[5 rows x 74 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv(\"loan.csv\", low_memory=False)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Remove columns with >90% missing values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "del_columns = []\n",
    "for col in data.columns:\n",
    "    if (data[col].isnull().sum() / len(data) * 100) > 90:\n",
    "        del_columns.append(col)\n",
    "\n",
    "data = data.drop(columns=del_columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Remove ID columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_columns = [\"id\", \"member_id\"]\n",
    "data = data.drop(columns=id_columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Explore individual columns (or column groups) for feature selection and extraction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Geographical features\n",
    "- addr_state\n",
    "- zip_code\n",
    "\n",
    "State and Zip code should be highly correlated. So, let's consider only the first 2 digitis of the Zip code and remove the state column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       addr_state zip_code\n",
      "count      887379   887379\n",
      "unique         51      935\n",
      "top            CA    945xx\n",
      "freq       129517     9770\n",
      "missing % addr_state: 0.0\n",
      "missing % zip_code: 0.0\n"
     ]
    }
   ],
   "source": [
    "feat_arr = [\"addr_state\", \"zip_code\"]\n",
    "print(data[feat_arr].describe())\n",
    "for col in feat_arr:\n",
    "    print(f\"missing % {col}:\", data[col].isnull().sum() / len(data) * 100)\n",
    "\n",
    "data[\"2_digit_zip\"] = data[\"zip_code\"].apply(lambda x: int(x[:2]))\n",
    "data = data.drop(columns=[\"addr_state\", \"zip_code\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Employment and income related features\n",
    "- annual_inc\n",
    "- verification_status\n",
    "- emp_length\n",
    "- emp_title\n",
    "\n",
    "Let's remove the employment title as we already have the annual income feature.\\\n",
    "The emp_length needs to be cleaned and converted to numerical feature.\\\n",
    "We fill the missing values using mean imputation and encode the verification_status feature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "         annual_inc\n",
      "count  8.873750e+05\n",
      "mean   7.502759e+04\n",
      "std    6.469830e+04\n",
      "min    0.000000e+00\n",
      "25%    4.500000e+04\n",
      "50%    6.500000e+04\n",
      "75%    9.000000e+04\n",
      "max    9.500000e+06\n",
      "       verification_status emp_length emp_title\n",
      "count               887379     842554    835917\n",
      "unique                   3         11    299271\n",
      "top        Source Verified  10+ years   Teacher\n",
      "freq                329558     291569     13469\n",
      "missing % annual_inc: 0.00045076568185634325\n",
      "missing % verification_status: 0.0\n",
      "missing % emp_length: 5.051392922302647\n",
      "missing % emp_title: 5.799325879922783\n"
     ]
    }
   ],
   "source": [
    "feat_arr_num = [\"annual_inc\"] \n",
    "feat_arr_cat = [\"verification_status\", \"emp_length\", \"emp_title\"]\n",
    "print(data[feat_arr_num].describe())\n",
    "print(data[feat_arr_cat].describe())\n",
    "for col in feat_arr_num+feat_arr_cat:\n",
    "    print(f\"missing % {col}:\", data[col].isnull().sum() / len(data) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def clean_emp_length(x):\n",
    "    if pd.isna(x): return None\n",
    "    if x.strip().lower()==\"< 1 year\": return 0\n",
    "    elif x.strip().lower()==\"1 year\": return 1\n",
    "    elif x.strip().lower()==\"10+ years\": return 11\n",
    "    else: return int(x.rstrip(\"years\").strip())\n",
    "\n",
    "data[\"emp_length\"] = data[\"emp_length\"].apply(clean_emp_length)\n",
    "data[\"emp_length\"] = data[\"emp_length\"].fillna(value=data[\"emp_length\"].mean())\n",
    "\n",
    "data[\"annual_inc\"] = data[\"annual_inc\"].fillna(value=data[\"annual_inc\"].mean())\n",
    "\n",
    "le = LabelEncoder()\n",
    "data['verification_status'] = le.fit_transform(data['verification_status'])\n",
    "\n",
    "data = data.drop(columns=[\"emp_title\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Metadata related to the loan application\n",
    "- purpose\n",
    "- desc\n",
    "- title\n",
    "- url\n",
    "- policy_code\n",
    "- application_type\n",
    "- initial_list_status\n",
    "- issue_d\n",
    "\n",
    "Url is not an important feature; hence, it is removed.\\\n",
    "The loan description and title are free-form text; hence, they are removed.\\\n",
    "The variances of policy_code and application_type are very low (~0); hence, they are also removed.\\\n",
    "We enode the purpose and initial_list_status and compute the time (years) since issue_d."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                   purpose initial_list_status   issue_d\n",
      "count               887379              887379    887379\n",
      "unique                  14                   2       103\n",
      "top     debt_consolidation                   f  Oct-2015\n",
      "freq                524215              456848     48631\n",
      "missing % purpose: 0.0\n",
      "missing % initial_list_status: 0.0\n",
      "missing % issue_d: 0.0\n"
     ]
    }
   ],
   "source": [
    "feat_arr_cat = [\"purpose\", \"initial_list_status\", \"issue_d\"]\n",
    "print(data[feat_arr_cat].describe())\n",
    "for col in feat_arr_cat:\n",
    "    print(f\"missing % {col}:\", data[col].isnull().sum() / len(data) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "data['purpose'] = le.fit_transform(data['purpose'])\n",
    "data['initial_list_status'] = le.fit_transform(data['initial_list_status'])\n",
    "\n",
    "def compute_time_since(x):\n",
    "    if pd.isna(x): return None\n",
    "    given_date = datetime.strptime(x, \"%b-%Y\")\n",
    "    current_date = datetime.now()\n",
    "    return (current_date - given_date).days / 365.25\n",
    "\n",
    "data[\"issue_d\"] = data[\"issue_d\"].apply(compute_time_since)\n",
    "\n",
    "data = data.drop(columns=[\"desc\", \"title\", \"url\", \"policy_code\", \"application_type\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### $$$ Loan, approved amount and payment related features\n",
    "- loan_amnt\n",
    "- int_rate\n",
    "- installment\n",
    "- funded_amnt\n",
    "- funded_amnt_inv\n",
    "- total_rec_int\n",
    "- total_rec_late_fee\n",
    "- total_rec_prncp\n",
    "- total_pymnt = total_rec_int + total_rec_prncp + total_rec_late_fee\n",
    "- total_pymnt_inv = total_rec_int + total_rec_prncp + total_rec_late_fee\n",
    "- last_pymnt_amnt\n",
    "- out_prncp = funded_amnt - total_rec_prncp\n",
    "- out_prncp_inv = funded_amnt_inv - total_rec_prncp\n",
    "- term\n",
    "- pymnt_plan\n",
    "- last_pymnt_d\n",
    "\n",
    "We note that some columns are derived from other columns. To avoid redundency, we remove the following columns - \\\n",
    "total_rec_int, total_rec_prncp, out_prncp and out_prncp_inv.\\\n",
    "We also remove pymnt_plan due to low variance.\\\n",
    "We enode the term and compute the time (years) since last_pymnt_d."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>loan_amnt</th>\n",
       "      <th>int_rate</th>\n",
       "      <th>installment</th>\n",
       "      <th>funded_amnt</th>\n",
       "      <th>funded_amnt_inv</th>\n",
       "      <th>total_pymnt</th>\n",
       "      <th>total_pymnt_inv</th>\n",
       "      <th>last_pymnt_amnt</th>\n",
       "      <th>total_rec_late_fee</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887379.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>14755.264605</td>\n",
       "      <td>13.246740</td>\n",
       "      <td>436.717127</td>\n",
       "      <td>14741.877625</td>\n",
       "      <td>14702.464383</td>\n",
       "      <td>7558.826684</td>\n",
       "      <td>7521.160306</td>\n",
       "      <td>2164.145585</td>\n",
       "      <td>0.396692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>8435.455601</td>\n",
       "      <td>4.381867</td>\n",
       "      <td>244.186593</td>\n",
       "      <td>8429.897657</td>\n",
       "      <td>8442.106732</td>\n",
       "      <td>7871.243336</td>\n",
       "      <td>7846.108604</td>\n",
       "      <td>4794.783233</td>\n",
       "      <td>4.087825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>500.000000</td>\n",
       "      <td>5.320000</td>\n",
       "      <td>15.670000</td>\n",
       "      <td>500.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>8000.000000</td>\n",
       "      <td>9.990000</td>\n",
       "      <td>260.705000</td>\n",
       "      <td>8000.000000</td>\n",
       "      <td>8000.000000</td>\n",
       "      <td>1914.590000</td>\n",
       "      <td>1899.700000</td>\n",
       "      <td>280.170000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>13000.000000</td>\n",
       "      <td>12.990000</td>\n",
       "      <td>382.550000</td>\n",
       "      <td>13000.000000</td>\n",
       "      <td>13000.000000</td>\n",
       "      <td>4894.999117</td>\n",
       "      <td>4862.010000</td>\n",
       "      <td>462.780000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>20000.000000</td>\n",
       "      <td>16.200000</td>\n",
       "      <td>572.600000</td>\n",
       "      <td>20000.000000</td>\n",
       "      <td>20000.000000</td>\n",
       "      <td>10616.814231</td>\n",
       "      <td>10565.690000</td>\n",
       "      <td>831.245000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>35000.000000</td>\n",
       "      <td>28.990000</td>\n",
       "      <td>1445.460000</td>\n",
       "      <td>35000.000000</td>\n",
       "      <td>35000.000000</td>\n",
       "      <td>57777.579870</td>\n",
       "      <td>57777.580000</td>\n",
       "      <td>36475.590000</td>\n",
       "      <td>358.680000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           loan_amnt       int_rate    installment    funded_amnt  \\\n",
       "count  887379.000000  887379.000000  887379.000000  887379.000000   \n",
       "mean    14755.264605      13.246740     436.717127   14741.877625   \n",
       "std      8435.455601       4.381867     244.186593    8429.897657   \n",
       "min       500.000000       5.320000      15.670000     500.000000   \n",
       "25%      8000.000000       9.990000     260.705000    8000.000000   \n",
       "50%     13000.000000      12.990000     382.550000   13000.000000   \n",
       "75%     20000.000000      16.200000     572.600000   20000.000000   \n",
       "max     35000.000000      28.990000    1445.460000   35000.000000   \n",
       "\n",
       "       funded_amnt_inv    total_pymnt  total_pymnt_inv  last_pymnt_amnt  \\\n",
       "count    887379.000000  887379.000000    887379.000000    887379.000000   \n",
       "mean      14702.464383    7558.826684      7521.160306      2164.145585   \n",
       "std        8442.106732    7871.243336      7846.108604      4794.783233   \n",
       "min           0.000000       0.000000         0.000000         0.000000   \n",
       "25%        8000.000000    1914.590000      1899.700000       280.170000   \n",
       "50%       13000.000000    4894.999117      4862.010000       462.780000   \n",
       "75%       20000.000000   10616.814231     10565.690000       831.245000   \n",
       "max       35000.000000   57777.579870     57777.580000     36475.590000   \n",
       "\n",
       "       total_rec_late_fee  \n",
       "count       887379.000000  \n",
       "mean             0.396692  \n",
       "std              4.087825  \n",
       "min              0.000000  \n",
       "25%              0.000000  \n",
       "50%              0.000000  \n",
       "75%              0.000000  \n",
       "max            358.680000  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feat_arr_num = [\"loan_amnt\", \"int_rate\", \"installment\", \"funded_amnt\", \"funded_amnt_inv\", \n",
    "\"total_pymnt\", \"total_pymnt_inv\", \"last_pymnt_amnt\", \"total_rec_late_fee\"]\n",
    "data[feat_arr_num].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>term</th>\n",
       "      <th>last_pymnt_d</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>887379</td>\n",
       "      <td>869720</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unique</th>\n",
       "      <td>2</td>\n",
       "      <td>98</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>top</th>\n",
       "      <td>36 months</td>\n",
       "      <td>Jan-2016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>freq</th>\n",
       "      <td>621125</td>\n",
       "      <td>470150</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              term last_pymnt_d\n",
       "count       887379       869720\n",
       "unique           2           98\n",
       "top      36 months     Jan-2016\n",
       "freq        621125       470150"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feat_arr_cat = [\"term\", \"last_pymnt_d\"]\n",
    "data[feat_arr_cat].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "missing % term: 0.0\n",
      "missing % last_pymnt_d: 1.9900177939752912\n",
      "missing % loan_amnt: 0.0\n",
      "missing % int_rate: 0.0\n",
      "missing % installment: 0.0\n",
      "missing % funded_amnt: 0.0\n",
      "missing % funded_amnt_inv: 0.0\n",
      "missing % total_pymnt: 0.0\n",
      "missing % total_pymnt_inv: 0.0\n",
      "missing % last_pymnt_amnt: 0.0\n",
      "missing % total_rec_late_fee: 0.0\n"
     ]
    }
   ],
   "source": [
    "for col in feat_arr_cat+feat_arr_num:\n",
    "    print(f\"missing % {col}:\", data[col].isnull().sum() / len(data) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "data['term'] = le.fit_transform(data['term'])\n",
    "\n",
    "data[\"last_pymnt_d\"] = data[\"last_pymnt_d\"].apply(compute_time_since)\n",
    "data[\"last_pymnt_d\"] = data[\"last_pymnt_d\"].fillna(value=data[\"last_pymnt_d\"].mean())\n",
    "\n",
    "data = data.drop(columns=[\"total_rec_int\", \"total_rec_prncp\", \"out_prncp\", \"out_prncp_inv\", \"pymnt_plan\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Borrower's financial history related features\n",
    "- dti\n",
    "- earliest_cr_line\n",
    "- delinq_2yrs\n",
    "- collections_12_mths_ex_med\n",
    "- inq_last_6mths\n",
    "- total_rev_hi_lim\n",
    "- acc_now_delinq\n",
    "- tot_coll_amt\n",
    "- tot_cur_bal\n",
    "- last_credit_pull_d\n",
    "- open_acc\n",
    "- total_acc\n",
    "- mths_since_last_delinq\n",
    "- mths_since_last_major_derog\n",
    "- mths_since_last_record\n",
    "- pub_rec\n",
    "- revol_bal\n",
    "- revol_util\n",
    "\n",
    "We keep all the features. The missing values are filled using mean imputation.\\\n",
    "earliest_cr_line and last_credit_pull_d are converted to numerical features using the aforementioned method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>dti</th>\n",
       "      <th>delinq_2yrs</th>\n",
       "      <th>collections_12_mths_ex_med</th>\n",
       "      <th>inq_last_6mths</th>\n",
       "      <th>total_rev_hi_lim</th>\n",
       "      <th>acc_now_delinq</th>\n",
       "      <th>tot_coll_amt</th>\n",
       "      <th>tot_cur_bal</th>\n",
       "      <th>open_acc</th>\n",
       "      <th>total_acc</th>\n",
       "      <th>mths_since_last_delinq</th>\n",
       "      <th>mths_since_last_major_derog</th>\n",
       "      <th>mths_since_last_record</th>\n",
       "      <th>pub_rec</th>\n",
       "      <th>revol_bal</th>\n",
       "      <th>revol_util</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>887379.000000</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>887234.000000</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>8.171030e+05</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>8.171030e+05</td>\n",
       "      <td>8.171030e+05</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>433067.000000</td>\n",
       "      <td>221703.000000</td>\n",
       "      <td>137053.000000</td>\n",
       "      <td>887350.000000</td>\n",
       "      <td>8.873790e+05</td>\n",
       "      <td>886877.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>18.157039</td>\n",
       "      <td>0.314442</td>\n",
       "      <td>0.014380</td>\n",
       "      <td>0.694623</td>\n",
       "      <td>3.206862e+04</td>\n",
       "      <td>0.004991</td>\n",
       "      <td>2.257026e+02</td>\n",
       "      <td>1.394582e+05</td>\n",
       "      <td>11.548469</td>\n",
       "      <td>25.268026</td>\n",
       "      <td>34.063798</td>\n",
       "      <td>44.104838</td>\n",
       "      <td>70.117903</td>\n",
       "      <td>0.195307</td>\n",
       "      <td>1.692079e+04</td>\n",
       "      <td>55.067693</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>17.190626</td>\n",
       "      <td>0.862244</td>\n",
       "      <td>0.134191</td>\n",
       "      <td>0.998448</td>\n",
       "      <td>3.749826e+04</td>\n",
       "      <td>0.077625</td>\n",
       "      <td>1.031137e+04</td>\n",
       "      <td>1.537500e+05</td>\n",
       "      <td>5.317313</td>\n",
       "      <td>11.840561</td>\n",
       "      <td>21.884940</td>\n",
       "      <td>22.179841</td>\n",
       "      <td>28.127914</td>\n",
       "      <td>0.582091</td>\n",
       "      <td>2.242679e+04</td>\n",
       "      <td>23.834344</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>11.910000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.390000e+04</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>2.985300e+04</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>17.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>27.000000</td>\n",
       "      <td>51.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>6.443000e+03</td>\n",
       "      <td>37.700000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>17.650000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.370000e+04</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>8.055900e+04</td>\n",
       "      <td>11.000000</td>\n",
       "      <td>24.000000</td>\n",
       "      <td>31.000000</td>\n",
       "      <td>44.000000</td>\n",
       "      <td>70.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.187500e+04</td>\n",
       "      <td>56.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>23.950000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>3.980000e+04</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>2.082050e+05</td>\n",
       "      <td>14.000000</td>\n",
       "      <td>32.000000</td>\n",
       "      <td>50.000000</td>\n",
       "      <td>61.000000</td>\n",
       "      <td>92.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.082900e+04</td>\n",
       "      <td>73.600000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>9999.000000</td>\n",
       "      <td>39.000000</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>33.000000</td>\n",
       "      <td>9.999999e+06</td>\n",
       "      <td>14.000000</td>\n",
       "      <td>9.152545e+06</td>\n",
       "      <td>8.000078e+06</td>\n",
       "      <td>90.000000</td>\n",
       "      <td>169.000000</td>\n",
       "      <td>188.000000</td>\n",
       "      <td>188.000000</td>\n",
       "      <td>129.000000</td>\n",
       "      <td>86.000000</td>\n",
       "      <td>2.904836e+06</td>\n",
       "      <td>892.300000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 dti    delinq_2yrs  collections_12_mths_ex_med  \\\n",
       "count  887379.000000  887350.000000               887234.000000   \n",
       "mean       18.157039       0.314442                    0.014380   \n",
       "std        17.190626       0.862244                    0.134191   \n",
       "min         0.000000       0.000000                    0.000000   \n",
       "25%        11.910000       0.000000                    0.000000   \n",
       "50%        17.650000       0.000000                    0.000000   \n",
       "75%        23.950000       0.000000                    0.000000   \n",
       "max      9999.000000      39.000000                   20.000000   \n",
       "\n",
       "       inq_last_6mths  total_rev_hi_lim  acc_now_delinq  tot_coll_amt  \\\n",
       "count   887350.000000      8.171030e+05   887350.000000  8.171030e+05   \n",
       "mean         0.694623      3.206862e+04        0.004991  2.257026e+02   \n",
       "std          0.998448      3.749826e+04        0.077625  1.031137e+04   \n",
       "min          0.000000      0.000000e+00        0.000000  0.000000e+00   \n",
       "25%          0.000000      1.390000e+04        0.000000  0.000000e+00   \n",
       "50%          0.000000      2.370000e+04        0.000000  0.000000e+00   \n",
       "75%          1.000000      3.980000e+04        0.000000  0.000000e+00   \n",
       "max         33.000000      9.999999e+06       14.000000  9.152545e+06   \n",
       "\n",
       "        tot_cur_bal       open_acc      total_acc  mths_since_last_delinq  \\\n",
       "count  8.171030e+05  887350.000000  887350.000000           433067.000000   \n",
       "mean   1.394582e+05      11.548469      25.268026               34.063798   \n",
       "std    1.537500e+05       5.317313      11.840561               21.884940   \n",
       "min    0.000000e+00       0.000000       1.000000                0.000000   \n",
       "25%    2.985300e+04       8.000000      17.000000               15.000000   \n",
       "50%    8.055900e+04      11.000000      24.000000               31.000000   \n",
       "75%    2.082050e+05      14.000000      32.000000               50.000000   \n",
       "max    8.000078e+06      90.000000     169.000000              188.000000   \n",
       "\n",
       "       mths_since_last_major_derog  mths_since_last_record        pub_rec  \\\n",
       "count                221703.000000           137053.000000  887350.000000   \n",
       "mean                     44.104838               70.117903       0.195307   \n",
       "std                      22.179841               28.127914       0.582091   \n",
       "min                       0.000000                0.000000       0.000000   \n",
       "25%                      27.000000               51.000000       0.000000   \n",
       "50%                      44.000000               70.000000       0.000000   \n",
       "75%                      61.000000               92.000000       0.000000   \n",
       "max                     188.000000              129.000000      86.000000   \n",
       "\n",
       "          revol_bal     revol_util  \n",
       "count  8.873790e+05  886877.000000  \n",
       "mean   1.692079e+04      55.067693  \n",
       "std    2.242679e+04      23.834344  \n",
       "min    0.000000e+00       0.000000  \n",
       "25%    6.443000e+03      37.700000  \n",
       "50%    1.187500e+04      56.000000  \n",
       "75%    2.082900e+04      73.600000  \n",
       "max    2.904836e+06     892.300000  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feat_arr_num = [\"dti\",\"earliest_cr_line\",\"delinq_2yrs\",\"collections_12_mths_ex_med\",\n",
    "\"inq_last_6mths\",\"total_rev_hi_lim\",\"acc_now_delinq\",\"tot_coll_amt\",\"tot_cur_bal\",\n",
    "\"last_credit_pull_d\",\"open_acc\",\"total_acc\",\"mths_since_last_delinq\",\n",
    "\"mths_since_last_major_derog\",\"mths_since_last_record\",\"pub_rec\",\"revol_bal\",\"revol_util\"]\n",
    "data[feat_arr_num].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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>earliest_cr_line</th>\n",
       "      <th>last_credit_pull_d</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>887350</td>\n",
       "      <td>887326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unique</th>\n",
       "      <td>697</td>\n",
       "      <td>103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>top</th>\n",
       "      <td>Aug-2001</td>\n",
       "      <td>Jan-2016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>freq</th>\n",
       "      <td>6659</td>\n",
       "      <td>730574</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       earliest_cr_line last_credit_pull_d\n",
       "count            887350             887326\n",
       "unique              697                103\n",
       "top            Aug-2001           Jan-2016\n",
       "freq               6659             730574"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feat_arr_cat = [\"earliest_cr_line\", \"last_credit_pull_d\"]\n",
    "data[feat_arr_cat].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "missing % earliest_cr_line: 0.0032680511934584885\n",
      "missing % last_credit_pull_d: 0.005972645284596547\n",
      "missing % dti: 0.0\n",
      "missing % earliest_cr_line: 0.0032680511934584885\n",
      "missing % delinq_2yrs: 0.0032680511934584885\n",
      "missing % collections_12_mths_ex_med: 0.016340255967292442\n",
      "missing % inq_last_6mths: 0.0032680511934584885\n",
      "missing % total_rev_hi_lim: 7.919502264534094\n",
      "missing % acc_now_delinq: 0.0032680511934584885\n",
      "missing % tot_coll_amt: 7.919502264534094\n",
      "missing % tot_cur_bal: 7.919502264534094\n",
      "missing % last_credit_pull_d: 0.005972645284596547\n",
      "missing % open_acc: 0.0032680511934584885\n",
      "missing % total_acc: 0.0032680511934584885\n",
      "missing % mths_since_last_delinq: 51.19706461387975\n",
      "missing % mths_since_last_major_derog: 75.01597400885078\n",
      "missing % mths_since_last_record: 84.55530275113566\n",
      "missing % pub_rec: 0.0032680511934584885\n",
      "missing % revol_bal: 0.0\n",
      "missing % revol_util: 0.05657109307297107\n"
     ]
    }
   ],
   "source": [
    "for col in feat_arr_cat+feat_arr_num:\n",
    "    print(f\"missing % {col}:\", data[col].isnull().sum() / len(data) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "data[\"earliest_cr_line\"] = data[\"earliest_cr_line\"].apply(compute_time_since)\n",
    "data[\"last_credit_pull_d\"] = data[\"last_credit_pull_d\"].apply(compute_time_since)\n",
    "\n",
    "data[feat_arr_num] = data[feat_arr_num].fillna(data[feat_arr_num].mean())\n",
    "data[feat_arr_cat] = data[feat_arr_cat].fillna(data[feat_arr_cat].mean())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Loan grade related features\n",
    "- grade\n",
    "- sub_grade\n",
    "\n",
    "We keep both features and encode them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "data['grade'] = le.fit_transform(data['grade'])\n",
    "data['sub_grade'] = le.fit_transform(data['sub_grade'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Home ownership related feature\n",
    "- home_ownership\n",
    "\n",
    "We keep this feature and encode it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "data['home_ownership'] = le.fit_transform(data['home_ownership'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Columns that could lead to data leakage\n",
    "- collection_recovery_fee (confirms that the loan has been charged off)\n",
    "- recoveries (confirms that the loan has been charged off)\n",
    "- next_pymnt_d (may reveal loan status if the date is NaN. I'm not sure about this though.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.drop(columns=[\"collection_recovery_fee\", \"recoveries\", \"next_pymnt_d\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Target class distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='count', ylabel='loan_status'>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.countplot(data=data, y=\"loan_status\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Binarize the target\n",
    "- 0: fully paid, current, etc. \n",
    "- 1: charged off, late, etc.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='count', ylabel='target'>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def binarize_target(loan_status):\n",
    "    cat_0_words = [\"paid\", \"current\"]\n",
    "    cat_1_words = [\"charged off\", \"late\", \"default\", \"grace period\"]\n",
    "    for word in cat_0_words:\n",
    "        if loan_status.lower().find(word)!=-1: return 0\n",
    "    for word in cat_1_words:\n",
    "        if loan_status.lower().find(word)!=-1: return 1\n",
    "    return -1\n",
    "\n",
    "data[\"target\"] = data[\"loan_status\"].apply(binarize_target)\n",
    "data = data[data[\"target\"]!=-1]\n",
    "sns.countplot(data=data, y=\"target\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Moleding and Hyperparameter Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 30/30 [27:58<00:00, 55.95s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'n_estimators': 300, 'criterion': 'gini', 'max_depth': 7}\n",
      "Best Score: {'Accuracy': 0.9196969007418195, 'Precision': 0.48950992630773565, 'Recall': 0.8941522618609783, 'F1-score': 0.6326636827313418, 'ROC AUC': 0.9079951821608439}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X = data.drop(columns=[\"loan_status\", \"target\"])\n",
    "y = data[[\"target\"]]\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=131543)\n",
    "\n",
    "# Grid search\n",
    "search_space = {\n",
    "    \"n_estimators\": [100, 200, 300, 400, 500],\n",
    "    \"criterion\": ['gini', 'entropy'],\n",
    "    \"max_depth\": [3, 5, 7]\n",
    "}\n",
    "\n",
    "import itertools\n",
    "parameter_combinations = list(itertools.product(\n",
    "    search_space['n_estimators'],\n",
    "    search_space['criterion'],\n",
    "    search_space['max_depth']\n",
    "))\n",
    "\n",
    "best_score = {\"F1-score\": 0}\n",
    "best_params = {}\n",
    "\n",
    "for n_estimators, criterion, max_depth in tqdm(parameter_combinations):\n",
    "    rf = RandomForestClassifier(\n",
    "        n_estimators=n_estimators,  \n",
    "        criterion=criterion,  \n",
    "        max_depth=max_depth,    \n",
    "        random_state=131543,\n",
    "        n_jobs=5,\n",
    "        class_weight=\"balanced\"\n",
    "    )\n",
    "    rf.fit(X_train, y_train.values.ravel())\n",
    "\n",
    "    curr_score = {\n",
    "        \"Accuracy\": accuracy_score(y_test, rf.predict(X_test)),\n",
    "        \"Precision\": precision_score(y_test, rf.predict(X_test)),\n",
    "        \"Recall\": recall_score(y_test, rf.predict(X_test)),\n",
    "        \"F1-score\": f1_score(y_test, rf.predict(X_test)),\n",
    "        \"ROC AUC\": roc_auc_score(y_test, rf.predict(X_test)),\n",
    "    }\n",
    "\n",
    "    if curr_score[\"F1-score\"] > best_score[\"F1-score\"]:\n",
    "        best_score = curr_score\n",
    "        best_params = {\n",
    "            'n_estimators': n_estimators,\n",
    "            'criterion': criterion,\n",
    "            'max_depth': max_depth\n",
    "        }\n",
    "\n",
    "print(\"Best Parameters:\", best_params)\n",
    "print(\"Best Score:\", best_score)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.13 ('base')",
   "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": "3.9.13"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "aceb42324b75a9bf78b98aa7346e5d8eff0fa8897468a0dab84ef13509829773"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}