Defining the problem.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Lending club underwriter - defining the problem\n",
    "The goal of this notebook is to define the target we are trying to predict and develop train/validation/testing sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The text.latex.preview rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The mathtext.fallback_to_cm rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: Support for setting the 'mathtext.fallback_to_cm' rcParam is deprecated since 3.3 and will be removed two minor releases later; use 'mathtext.fallback : 'cm' instead.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The validate_bool_maybe_none function was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The savefig.jpeg_quality rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The keymap.all_axes rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The animation.avconv_path rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n",
      "In C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\matplotlib\\mpl-data\\stylelib\\_classic_test.mplstyle: \n",
      "The animation.avconv_args rcparam was deprecated in Matplotlib 3.3 and will be removed two minor releases later.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.patches as mpatches\n",
    "from datetime import datetime\n",
    "\n",
    "np.random.seed(123) # set numpy random seed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\tysonp\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py:3147: DtypeWarning: Columns (0,19,49,59,118,129,130,131,134,135,136,139,145,146,147) have mixed types.Specify dtype option on import or set low_memory=False.\n",
      "  interactivity=interactivity, compiler=compiler, result=result)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(2260701, 151)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"accepted_2007_to_2018Q4.csv\")\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are 2.2M rows (each row corresponding to a unique loans) and 151 columns."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "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>emp_title</th>\n",
       "      <th>emp_length</th>\n",
       "      <th>home_ownership</th>\n",
       "      <th>annual_inc</th>\n",
       "      <th>verification_status</th>\n",
       "      <th>issue_d</th>\n",
       "      <th>loan_status</th>\n",
       "      <th>pymnt_plan</th>\n",
       "      <th>url</th>\n",
       "      <th>desc</th>\n",
       "      <th>purpose</th>\n",
       "      <th>title</th>\n",
       "      <th>zip_code</th>\n",
       "      <th>addr_state</th>\n",
       "      <th>dti</th>\n",
       "      <th>delinq_2yrs</th>\n",
       "      <th>earliest_cr_line</th>\n",
       "      <th>fico_range_low</th>\n",
       "      <th>fico_range_high</th>\n",
       "      <th>inq_last_6mths</th>\n",
       "      <th>mths_since_last_delinq</th>\n",
       "      <th>mths_since_last_record</th>\n",
       "      <th>open_acc</th>\n",
       "      <th>pub_rec</th>\n",
       "      <th>revol_bal</th>\n",
       "      <th>revol_util</th>\n",
       "      <th>total_acc</th>\n",
       "      <th>initial_list_status</th>\n",
       "      <th>out_prncp</th>\n",
       "      <th>out_prncp_inv</th>\n",
       "      <th>total_pymnt</th>\n",
       "      <th>total_pymnt_inv</th>\n",
       "      <th>total_rec_prncp</th>\n",
       "      <th>total_rec_int</th>\n",
       "      <th>total_rec_late_fee</th>\n",
       "      <th>recoveries</th>\n",
       "      <th>collection_recovery_fee</th>\n",
       "      <th>last_pymnt_d</th>\n",
       "      <th>last_pymnt_amnt</th>\n",
       "      <th>next_pymnt_d</th>\n",
       "      <th>last_credit_pull_d</th>\n",
       "      <th>last_fico_range_high</th>\n",
       "      <th>last_fico_range_low</th>\n",
       "      <th>collections_12_mths_ex_med</th>\n",
       "      <th>mths_since_last_major_derog</th>\n",
       "      <th>policy_code</th>\n",
       "      <th>application_type</th>\n",
       "      <th>annual_inc_joint</th>\n",
       "      <th>dti_joint</th>\n",
       "      <th>verification_status_joint</th>\n",
       "      <th>acc_now_delinq</th>\n",
       "      <th>tot_coll_amt</th>\n",
       "      <th>tot_cur_bal</th>\n",
       "      <th>open_acc_6m</th>\n",
       "      <th>open_act_il</th>\n",
       "      <th>open_il_12m</th>\n",
       "      <th>open_il_24m</th>\n",
       "      <th>mths_since_rcnt_il</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",
       "      <th>acc_open_past_24mths</th>\n",
       "      <th>avg_cur_bal</th>\n",
       "      <th>bc_open_to_buy</th>\n",
       "      <th>bc_util</th>\n",
       "      <th>chargeoff_within_12_mths</th>\n",
       "      <th>delinq_amnt</th>\n",
       "      <th>mo_sin_old_il_acct</th>\n",
       "      <th>mo_sin_old_rev_tl_op</th>\n",
       "      <th>mo_sin_rcnt_rev_tl_op</th>\n",
       "      <th>mo_sin_rcnt_tl</th>\n",
       "      <th>mort_acc</th>\n",
       "      <th>mths_since_recent_bc</th>\n",
       "      <th>mths_since_recent_bc_dlq</th>\n",
       "      <th>mths_since_recent_inq</th>\n",
       "      <th>mths_since_recent_revol_delinq</th>\n",
       "      <th>num_accts_ever_120_pd</th>\n",
       "      <th>num_actv_bc_tl</th>\n",
       "      <th>num_actv_rev_tl</th>\n",
       "      <th>num_bc_sats</th>\n",
       "      <th>num_bc_tl</th>\n",
       "      <th>num_il_tl</th>\n",
       "      <th>num_op_rev_tl</th>\n",
       "      <th>num_rev_accts</th>\n",
       "      <th>num_rev_tl_bal_gt_0</th>\n",
       "      <th>num_sats</th>\n",
       "      <th>num_tl_120dpd_2m</th>\n",
       "      <th>num_tl_30dpd</th>\n",
       "      <th>num_tl_90g_dpd_24m</th>\n",
       "      <th>num_tl_op_past_12m</th>\n",
       "      <th>pct_tl_nvr_dlq</th>\n",
       "      <th>percent_bc_gt_75</th>\n",
       "      <th>pub_rec_bankruptcies</th>\n",
       "      <th>tax_liens</th>\n",
       "      <th>tot_hi_cred_lim</th>\n",
       "      <th>total_bal_ex_mort</th>\n",
       "      <th>total_bc_limit</th>\n",
       "      <th>total_il_high_credit_limit</th>\n",
       "      <th>revol_bal_joint</th>\n",
       "      <th>sec_app_fico_range_low</th>\n",
       "      <th>sec_app_fico_range_high</th>\n",
       "      <th>sec_app_earliest_cr_line</th>\n",
       "      <th>sec_app_inq_last_6mths</th>\n",
       "      <th>sec_app_mort_acc</th>\n",
       "      <th>sec_app_open_acc</th>\n",
       "      <th>sec_app_revol_util</th>\n",
       "      <th>sec_app_open_act_il</th>\n",
       "      <th>sec_app_num_rev_accts</th>\n",
       "      <th>sec_app_chargeoff_within_12_mths</th>\n",
       "      <th>sec_app_collections_12_mths_ex_med</th>\n",
       "      <th>sec_app_mths_since_last_major_derog</th>\n",
       "      <th>hardship_flag</th>\n",
       "      <th>hardship_type</th>\n",
       "      <th>hardship_reason</th>\n",
       "      <th>hardship_status</th>\n",
       "      <th>deferral_term</th>\n",
       "      <th>hardship_amount</th>\n",
       "      <th>hardship_start_date</th>\n",
       "      <th>hardship_end_date</th>\n",
       "      <th>payment_plan_start_date</th>\n",
       "      <th>hardship_length</th>\n",
       "      <th>hardship_dpd</th>\n",
       "      <th>hardship_loan_status</th>\n",
       "      <th>orig_projected_additional_accrued_interest</th>\n",
       "      <th>hardship_payoff_balance_amount</th>\n",
       "      <th>hardship_last_payment_amount</th>\n",
       "      <th>disbursement_method</th>\n",
       "      <th>debt_settlement_flag</th>\n",
       "      <th>debt_settlement_flag_date</th>\n",
       "      <th>settlement_status</th>\n",
       "      <th>settlement_date</th>\n",
       "      <th>settlement_amount</th>\n",
       "      <th>settlement_percentage</th>\n",
       "      <th>settlement_term</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>68407277</td>\n",
       "      <td>NaN</td>\n",
       "      <td>3600.0</td>\n",
       "      <td>3600.0</td>\n",
       "      <td>3600.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>13.99</td>\n",
       "      <td>123.03</td>\n",
       "      <td>C</td>\n",
       "      <td>C4</td>\n",
       "      <td>leadman</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>55000.0</td>\n",
       "      <td>Not Verified</td>\n",
       "      <td>Dec-2015</td>\n",
       "      <td>Fully Paid</td>\n",
       "      <td>n</td>\n",
       "      <td>https://lendingclub.com/browse/loanDetail.acti...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>debt_consolidation</td>\n",
       "      <td>Debt consolidation</td>\n",
       "      <td>190xx</td>\n",
       "      <td>PA</td>\n",
       "      <td>5.91</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Aug-2003</td>\n",
       "      <td>675.0</td>\n",
       "      <td>679.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2765.0</td>\n",
       "      <td>29.7</td>\n",
       "      <td>13.0</td>\n",
       "      <td>w</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>4421.723917</td>\n",
       "      <td>4421.72</td>\n",
       "      <td>3600.00</td>\n",
       "      <td>821.72</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Jan-2019</td>\n",
       "      <td>122.67</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Mar-2019</td>\n",
       "      <td>564.0</td>\n",
       "      <td>560.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Individual</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>722.0</td>\n",
       "      <td>144904.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>4981.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>722.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>9300.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>20701.0</td>\n",
       "      <td>1506.0</td>\n",
       "      <td>37.2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>148.0</td>\n",
       "      <td>128.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>69.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>69.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>76.9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>178050.0</td>\n",
       "      <td>7746.0</td>\n",
       "      <td>2400.0</td>\n",
       "      <td>13734.0</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>N</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Cash</td>\n",
       "      <td>N</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>68355089</td>\n",
       "      <td>NaN</td>\n",
       "      <td>24700.0</td>\n",
       "      <td>24700.0</td>\n",
       "      <td>24700.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>11.99</td>\n",
       "      <td>820.28</td>\n",
       "      <td>C</td>\n",
       "      <td>C1</td>\n",
       "      <td>Engineer</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>65000.0</td>\n",
       "      <td>Not Verified</td>\n",
       "      <td>Dec-2015</td>\n",
       "      <td>Fully Paid</td>\n",
       "      <td>n</td>\n",
       "      <td>https://lendingclub.com/browse/loanDetail.acti...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>small_business</td>\n",
       "      <td>Business</td>\n",
       "      <td>577xx</td>\n",
       "      <td>SD</td>\n",
       "      <td>16.06</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Dec-1999</td>\n",
       "      <td>715.0</td>\n",
       "      <td>719.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>22.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>21470.0</td>\n",
       "      <td>19.2</td>\n",
       "      <td>38.0</td>\n",
       "      <td>w</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>25679.660000</td>\n",
       "      <td>25679.66</td>\n",
       "      <td>24700.00</td>\n",
       "      <td>979.66</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Jun-2016</td>\n",
       "      <td>926.35</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Mar-2019</td>\n",
       "      <td>699.0</td>\n",
       "      <td>695.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Individual</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>204396.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>18005.0</td>\n",
       "      <td>73.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>6472.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>111800.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>9733.0</td>\n",
       "      <td>57830.0</td>\n",
       "      <td>27.1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>113.0</td>\n",
       "      <td>192.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>20.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>97.4</td>\n",
       "      <td>7.7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>314017.0</td>\n",
       "      <td>39475.0</td>\n",
       "      <td>79300.0</td>\n",
       "      <td>24667.0</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>N</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Cash</td>\n",
       "      <td>N</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>68341763</td>\n",
       "      <td>NaN</td>\n",
       "      <td>20000.0</td>\n",
       "      <td>20000.0</td>\n",
       "      <td>20000.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>10.78</td>\n",
       "      <td>432.66</td>\n",
       "      <td>B</td>\n",
       "      <td>B4</td>\n",
       "      <td>truck driver</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>63000.0</td>\n",
       "      <td>Not Verified</td>\n",
       "      <td>Dec-2015</td>\n",
       "      <td>Fully Paid</td>\n",
       "      <td>n</td>\n",
       "      <td>https://lendingclub.com/browse/loanDetail.acti...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>home_improvement</td>\n",
       "      <td>NaN</td>\n",
       "      <td>605xx</td>\n",
       "      <td>IL</td>\n",
       "      <td>10.78</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Aug-2000</td>\n",
       "      <td>695.0</td>\n",
       "      <td>699.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7869.0</td>\n",
       "      <td>56.2</td>\n",
       "      <td>18.0</td>\n",
       "      <td>w</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>22705.924294</td>\n",
       "      <td>22705.92</td>\n",
       "      <td>20000.00</td>\n",
       "      <td>2705.92</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Jun-2017</td>\n",
       "      <td>15813.30</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Mar-2019</td>\n",
       "      <td>704.0</td>\n",
       "      <td>700.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Joint App</td>\n",
       "      <td>71000.0</td>\n",
       "      <td>13.85</td>\n",
       "      <td>Not Verified</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>189699.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>10827.0</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2081.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>14000.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>31617.0</td>\n",
       "      <td>2737.0</td>\n",
       "      <td>55.9</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>125.0</td>\n",
       "      <td>184.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>101.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>10.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>218418.0</td>\n",
       "      <td>18696.0</td>\n",
       "      <td>6200.0</td>\n",
       "      <td>14877.0</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>N</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Cash</td>\n",
       "      <td>N</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>66310712</td>\n",
       "      <td>NaN</td>\n",
       "      <td>35000.0</td>\n",
       "      <td>35000.0</td>\n",
       "      <td>35000.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>14.85</td>\n",
       "      <td>829.90</td>\n",
       "      <td>C</td>\n",
       "      <td>C5</td>\n",
       "      <td>Information Systems Officer</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>110000.0</td>\n",
       "      <td>Source Verified</td>\n",
       "      <td>Dec-2015</td>\n",
       "      <td>Current</td>\n",
       "      <td>n</td>\n",
       "      <td>https://lendingclub.com/browse/loanDetail.acti...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>debt_consolidation</td>\n",
       "      <td>Debt consolidation</td>\n",
       "      <td>076xx</td>\n",
       "      <td>NJ</td>\n",
       "      <td>17.06</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Sep-2008</td>\n",
       "      <td>785.0</td>\n",
       "      <td>789.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7802.0</td>\n",
       "      <td>11.6</td>\n",
       "      <td>17.0</td>\n",
       "      <td>w</td>\n",
       "      <td>15897.65</td>\n",
       "      <td>15897.65</td>\n",
       "      <td>31464.010000</td>\n",
       "      <td>31464.01</td>\n",
       "      <td>19102.35</td>\n",
       "      <td>12361.66</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Feb-2019</td>\n",
       "      <td>829.90</td>\n",
       "      <td>Apr-2019</td>\n",
       "      <td>Mar-2019</td>\n",
       "      <td>679.0</td>\n",
       "      <td>675.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Individual</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>301500.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>12609.0</td>\n",
       "      <td>70.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>6987.0</td>\n",
       "      <td>45.0</td>\n",
       "      <td>67300.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>23192.0</td>\n",
       "      <td>54962.0</td>\n",
       "      <td>12.1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>381215.0</td>\n",
       "      <td>52226.0</td>\n",
       "      <td>62500.0</td>\n",
       "      <td>18000.0</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>N</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Cash</td>\n",
       "      <td>N</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>68476807</td>\n",
       "      <td>NaN</td>\n",
       "      <td>10400.0</td>\n",
       "      <td>10400.0</td>\n",
       "      <td>10400.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>22.45</td>\n",
       "      <td>289.91</td>\n",
       "      <td>F</td>\n",
       "      <td>F1</td>\n",
       "      <td>Contract Specialist</td>\n",
       "      <td>3 years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>104433.0</td>\n",
       "      <td>Source Verified</td>\n",
       "      <td>Dec-2015</td>\n",
       "      <td>Fully Paid</td>\n",
       "      <td>n</td>\n",
       "      <td>https://lendingclub.com/browse/loanDetail.acti...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>major_purchase</td>\n",
       "      <td>Major purchase</td>\n",
       "      <td>174xx</td>\n",
       "      <td>PA</td>\n",
       "      <td>25.37</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Jun-1998</td>\n",
       "      <td>695.0</td>\n",
       "      <td>699.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>21929.0</td>\n",
       "      <td>64.5</td>\n",
       "      <td>35.0</td>\n",
       "      <td>w</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>11740.500000</td>\n",
       "      <td>11740.50</td>\n",
       "      <td>10400.00</td>\n",
       "      <td>1340.50</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>Jul-2016</td>\n",
       "      <td>10128.96</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Mar-2018</td>\n",
       "      <td>704.0</td>\n",
       "      <td>700.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>Individual</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>331730.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>73839.0</td>\n",
       "      <td>84.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>9702.0</td>\n",
       "      <td>78.0</td>\n",
       "      <td>34000.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>27644.0</td>\n",
       "      <td>4567.0</td>\n",
       "      <td>77.5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>128.0</td>\n",
       "      <td>210.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>96.6</td>\n",
       "      <td>60.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>439570.0</td>\n",
       "      <td>95768.0</td>\n",
       "      <td>20300.0</td>\n",
       "      <td>88097.0</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>N</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",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Cash</td>\n",
       "      <td>N</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",
       "</div>"
      ],
      "text/plain": [
       "         id  member_id  loan_amnt  funded_amnt  funded_amnt_inv        term  \\\n",
       "0  68407277        NaN     3600.0       3600.0           3600.0   36 months   \n",
       "1  68355089        NaN    24700.0      24700.0          24700.0   36 months   \n",
       "2  68341763        NaN    20000.0      20000.0          20000.0   60 months   \n",
       "3  66310712        NaN    35000.0      35000.0          35000.0   60 months   \n",
       "4  68476807        NaN    10400.0      10400.0          10400.0   60 months   \n",
       "\n",
       "   int_rate  installment grade sub_grade                    emp_title  \\\n",
       "0     13.99       123.03     C        C4                      leadman   \n",
       "1     11.99       820.28     C        C1                     Engineer   \n",
       "2     10.78       432.66     B        B4                 truck driver   \n",
       "3     14.85       829.90     C        C5  Information Systems Officer   \n",
       "4     22.45       289.91     F        F1          Contract Specialist   \n",
       "\n",
       "  emp_length home_ownership  annual_inc verification_status   issue_d  \\\n",
       "0  10+ years       MORTGAGE     55000.0        Not Verified  Dec-2015   \n",
       "1  10+ years       MORTGAGE     65000.0        Not Verified  Dec-2015   \n",
       "2  10+ years       MORTGAGE     63000.0        Not Verified  Dec-2015   \n",
       "3  10+ years       MORTGAGE    110000.0     Source Verified  Dec-2015   \n",
       "4    3 years       MORTGAGE    104433.0     Source Verified  Dec-2015   \n",
       "\n",
       "  loan_status pymnt_plan                                                url  \\\n",
       "0  Fully Paid          n  https://lendingclub.com/browse/loanDetail.acti...   \n",
       "1  Fully Paid          n  https://lendingclub.com/browse/loanDetail.acti...   \n",
       "2  Fully Paid          n  https://lendingclub.com/browse/loanDetail.acti...   \n",
       "3     Current          n  https://lendingclub.com/browse/loanDetail.acti...   \n",
       "4  Fully Paid          n  https://lendingclub.com/browse/loanDetail.acti...   \n",
       "\n",
       "  desc             purpose               title zip_code addr_state    dti  \\\n",
       "0  NaN  debt_consolidation  Debt consolidation    190xx         PA   5.91   \n",
       "1  NaN      small_business            Business    577xx         SD  16.06   \n",
       "2  NaN    home_improvement                 NaN    605xx         IL  10.78   \n",
       "3  NaN  debt_consolidation  Debt consolidation    076xx         NJ  17.06   \n",
       "4  NaN      major_purchase      Major purchase    174xx         PA  25.37   \n",
       "\n",
       "   delinq_2yrs earliest_cr_line  fico_range_low  fico_range_high  \\\n",
       "0          0.0         Aug-2003           675.0            679.0   \n",
       "1          1.0         Dec-1999           715.0            719.0   \n",
       "2          0.0         Aug-2000           695.0            699.0   \n",
       "3          0.0         Sep-2008           785.0            789.0   \n",
       "4          1.0         Jun-1998           695.0            699.0   \n",
       "\n",
       "   inq_last_6mths  mths_since_last_delinq  mths_since_last_record  open_acc  \\\n",
       "0             1.0                    30.0                     NaN       7.0   \n",
       "1             4.0                     6.0                     NaN      22.0   \n",
       "2             0.0                     NaN                     NaN       6.0   \n",
       "3             0.0                     NaN                     NaN      13.0   \n",
       "4             3.0                    12.0                     NaN      12.0   \n",
       "\n",
       "   pub_rec  revol_bal  revol_util  total_acc initial_list_status  out_prncp  \\\n",
       "0      0.0     2765.0        29.7       13.0                   w       0.00   \n",
       "1      0.0    21470.0        19.2       38.0                   w       0.00   \n",
       "2      0.0     7869.0        56.2       18.0                   w       0.00   \n",
       "3      0.0     7802.0        11.6       17.0                   w   15897.65   \n",
       "4      0.0    21929.0        64.5       35.0                   w       0.00   \n",
       "\n",
       "   out_prncp_inv   total_pymnt  total_pymnt_inv  total_rec_prncp  \\\n",
       "0           0.00   4421.723917          4421.72          3600.00   \n",
       "1           0.00  25679.660000         25679.66         24700.00   \n",
       "2           0.00  22705.924294         22705.92         20000.00   \n",
       "3       15897.65  31464.010000         31464.01         19102.35   \n",
       "4           0.00  11740.500000         11740.50         10400.00   \n",
       "\n",
       "   total_rec_int  total_rec_late_fee  recoveries  collection_recovery_fee  \\\n",
       "0         821.72                 0.0         0.0                      0.0   \n",
       "1         979.66                 0.0         0.0                      0.0   \n",
       "2        2705.92                 0.0         0.0                      0.0   \n",
       "3       12361.66                 0.0         0.0                      0.0   \n",
       "4        1340.50                 0.0         0.0                      0.0   \n",
       "\n",
       "  last_pymnt_d  last_pymnt_amnt next_pymnt_d last_credit_pull_d  \\\n",
       "0     Jan-2019           122.67          NaN           Mar-2019   \n",
       "1     Jun-2016           926.35          NaN           Mar-2019   \n",
       "2     Jun-2017         15813.30          NaN           Mar-2019   \n",
       "3     Feb-2019           829.90     Apr-2019           Mar-2019   \n",
       "4     Jul-2016         10128.96          NaN           Mar-2018   \n",
       "\n",
       "   last_fico_range_high  last_fico_range_low  collections_12_mths_ex_med  \\\n",
       "0                 564.0                560.0                         0.0   \n",
       "1                 699.0                695.0                         0.0   \n",
       "2                 704.0                700.0                         0.0   \n",
       "3                 679.0                675.0                         0.0   \n",
       "4                 704.0                700.0                         0.0   \n",
       "\n",
       "   mths_since_last_major_derog  policy_code application_type  \\\n",
       "0                         30.0          1.0       Individual   \n",
       "1                          NaN          1.0       Individual   \n",
       "2                          NaN          1.0        Joint App   \n",
       "3                          NaN          1.0       Individual   \n",
       "4                          NaN          1.0       Individual   \n",
       "\n",
       "   annual_inc_joint  dti_joint verification_status_joint  acc_now_delinq  \\\n",
       "0               NaN        NaN                       NaN             0.0   \n",
       "1               NaN        NaN                       NaN             0.0   \n",
       "2           71000.0      13.85              Not Verified             0.0   \n",
       "3               NaN        NaN                       NaN             0.0   \n",
       "4               NaN        NaN                       NaN             0.0   \n",
       "\n",
       "   tot_coll_amt  tot_cur_bal  open_acc_6m  open_act_il  open_il_12m  \\\n",
       "0         722.0     144904.0          2.0          2.0          0.0   \n",
       "1           0.0     204396.0          1.0          1.0          0.0   \n",
       "2           0.0     189699.0          0.0          1.0          0.0   \n",
       "3           0.0     301500.0          1.0          1.0          0.0   \n",
       "4           0.0     331730.0          1.0          3.0          0.0   \n",
       "\n",
       "   open_il_24m  mths_since_rcnt_il  total_bal_il  il_util  open_rv_12m  \\\n",
       "0          1.0                21.0        4981.0     36.0          3.0   \n",
       "1          1.0                19.0       18005.0     73.0          2.0   \n",
       "2          4.0                19.0       10827.0     73.0          0.0   \n",
       "3          1.0                23.0       12609.0     70.0          1.0   \n",
       "4          3.0                14.0       73839.0     84.0          4.0   \n",
       "\n",
       "   open_rv_24m  max_bal_bc  all_util  total_rev_hi_lim  inq_fi  total_cu_tl  \\\n",
       "0          3.0       722.0      34.0            9300.0     3.0          1.0   \n",
       "1          3.0      6472.0      29.0          111800.0     0.0          0.0   \n",
       "2          2.0      2081.0      65.0           14000.0     2.0          5.0   \n",
       "3          1.0      6987.0      45.0           67300.0     0.0          1.0   \n",
       "4          7.0      9702.0      78.0           34000.0     2.0          1.0   \n",
       "\n",
       "   inq_last_12m  acc_open_past_24mths  avg_cur_bal  bc_open_to_buy  bc_util  \\\n",
       "0           4.0                   4.0      20701.0          1506.0     37.2   \n",
       "1           6.0                   4.0       9733.0         57830.0     27.1   \n",
       "2           1.0                   6.0      31617.0          2737.0     55.9   \n",
       "3           0.0                   2.0      23192.0         54962.0     12.1   \n",
       "4           3.0                  10.0      27644.0          4567.0     77.5   \n",
       "\n",
       "   chargeoff_within_12_mths  delinq_amnt  mo_sin_old_il_acct  \\\n",
       "0                       0.0          0.0               148.0   \n",
       "1                       0.0          0.0               113.0   \n",
       "2                       0.0          0.0               125.0   \n",
       "3                       0.0          0.0                36.0   \n",
       "4                       0.0          0.0               128.0   \n",
       "\n",
       "   mo_sin_old_rev_tl_op  mo_sin_rcnt_rev_tl_op  mo_sin_rcnt_tl  mort_acc  \\\n",
       "0                 128.0                    3.0             3.0       1.0   \n",
       "1                 192.0                    2.0             2.0       4.0   \n",
       "2                 184.0                   14.0            14.0       5.0   \n",
       "3                  87.0                    2.0             2.0       1.0   \n",
       "4                 210.0                    4.0             4.0       6.0   \n",
       "\n",
       "   mths_since_recent_bc  mths_since_recent_bc_dlq  mths_since_recent_inq  \\\n",
       "0                   4.0                      69.0                    4.0   \n",
       "1                   2.0                       NaN                    0.0   \n",
       "2                 101.0                       NaN                   10.0   \n",
       "3                   2.0                       NaN                    NaN   \n",
       "4                   4.0                      12.0                    1.0   \n",
       "\n",
       "   mths_since_recent_revol_delinq  num_accts_ever_120_pd  num_actv_bc_tl  \\\n",
       "0                            69.0                    2.0             2.0   \n",
       "1                             6.0                    0.0             5.0   \n",
       "2                             NaN                    0.0             2.0   \n",
       "3                             NaN                    0.0             4.0   \n",
       "4                            12.0                    0.0             4.0   \n",
       "\n",
       "   num_actv_rev_tl  num_bc_sats  num_bc_tl  num_il_tl  num_op_rev_tl  \\\n",
       "0              4.0          2.0        5.0        3.0            4.0   \n",
       "1              5.0         13.0       17.0        6.0           20.0   \n",
       "2              3.0          2.0        4.0        6.0            4.0   \n",
       "3              5.0          8.0       10.0        2.0           10.0   \n",
       "4              6.0          5.0        9.0       10.0            7.0   \n",
       "\n",
       "   num_rev_accts  num_rev_tl_bal_gt_0  num_sats  num_tl_120dpd_2m  \\\n",
       "0            9.0                  4.0       7.0               0.0   \n",
       "1           27.0                  5.0      22.0               0.0   \n",
       "2            7.0                  3.0       6.0               0.0   \n",
       "3           13.0                  5.0      13.0               0.0   \n",
       "4           19.0                  6.0      12.0               0.0   \n",
       "\n",
       "   num_tl_30dpd  num_tl_90g_dpd_24m  num_tl_op_past_12m  pct_tl_nvr_dlq  \\\n",
       "0           0.0                 0.0                 3.0            76.9   \n",
       "1           0.0                 0.0                 2.0            97.4   \n",
       "2           0.0                 0.0                 0.0           100.0   \n",
       "3           0.0                 0.0                 1.0           100.0   \n",
       "4           0.0                 0.0                 4.0            96.6   \n",
       "\n",
       "   percent_bc_gt_75  pub_rec_bankruptcies  tax_liens  tot_hi_cred_lim  \\\n",
       "0               0.0                   0.0        0.0         178050.0   \n",
       "1               7.7                   0.0        0.0         314017.0   \n",
       "2              50.0                   0.0        0.0         218418.0   \n",
       "3               0.0                   0.0        0.0         381215.0   \n",
       "4              60.0                   0.0        0.0         439570.0   \n",
       "\n",
       "   total_bal_ex_mort  total_bc_limit  total_il_high_credit_limit  \\\n",
       "0             7746.0          2400.0                     13734.0   \n",
       "1            39475.0         79300.0                     24667.0   \n",
       "2            18696.0          6200.0                     14877.0   \n",
       "3            52226.0         62500.0                     18000.0   \n",
       "4            95768.0         20300.0                     88097.0   \n",
       "\n",
       "   revol_bal_joint  sec_app_fico_range_low  sec_app_fico_range_high  \\\n",
       "0              NaN                     NaN                      NaN   \n",
       "1              NaN                     NaN                      NaN   \n",
       "2              NaN                     NaN                      NaN   \n",
       "3              NaN                     NaN                      NaN   \n",
       "4              NaN                     NaN                      NaN   \n",
       "\n",
       "  sec_app_earliest_cr_line  sec_app_inq_last_6mths  sec_app_mort_acc  \\\n",
       "0                      NaN                     NaN               NaN   \n",
       "1                      NaN                     NaN               NaN   \n",
       "2                      NaN                     NaN               NaN   \n",
       "3                      NaN                     NaN               NaN   \n",
       "4                      NaN                     NaN               NaN   \n",
       "\n",
       "   sec_app_open_acc  sec_app_revol_util  sec_app_open_act_il  \\\n",
       "0               NaN                 NaN                  NaN   \n",
       "1               NaN                 NaN                  NaN   \n",
       "2               NaN                 NaN                  NaN   \n",
       "3               NaN                 NaN                  NaN   \n",
       "4               NaN                 NaN                  NaN   \n",
       "\n",
       "   sec_app_num_rev_accts  sec_app_chargeoff_within_12_mths  \\\n",
       "0                    NaN                               NaN   \n",
       "1                    NaN                               NaN   \n",
       "2                    NaN                               NaN   \n",
       "3                    NaN                               NaN   \n",
       "4                    NaN                               NaN   \n",
       "\n",
       "   sec_app_collections_12_mths_ex_med  sec_app_mths_since_last_major_derog  \\\n",
       "0                                 NaN                                  NaN   \n",
       "1                                 NaN                                  NaN   \n",
       "2                                 NaN                                  NaN   \n",
       "3                                 NaN                                  NaN   \n",
       "4                                 NaN                                  NaN   \n",
       "\n",
       "  hardship_flag hardship_type hardship_reason hardship_status  deferral_term  \\\n",
       "0             N           NaN             NaN             NaN            NaN   \n",
       "1             N           NaN             NaN             NaN            NaN   \n",
       "2             N           NaN             NaN             NaN            NaN   \n",
       "3             N           NaN             NaN             NaN            NaN   \n",
       "4             N           NaN             NaN             NaN            NaN   \n",
       "\n",
       "   hardship_amount hardship_start_date hardship_end_date  \\\n",
       "0              NaN                 NaN               NaN   \n",
       "1              NaN                 NaN               NaN   \n",
       "2              NaN                 NaN               NaN   \n",
       "3              NaN                 NaN               NaN   \n",
       "4              NaN                 NaN               NaN   \n",
       "\n",
       "  payment_plan_start_date  hardship_length  hardship_dpd hardship_loan_status  \\\n",
       "0                     NaN              NaN           NaN                  NaN   \n",
       "1                     NaN              NaN           NaN                  NaN   \n",
       "2                     NaN              NaN           NaN                  NaN   \n",
       "3                     NaN              NaN           NaN                  NaN   \n",
       "4                     NaN              NaN           NaN                  NaN   \n",
       "\n",
       "   orig_projected_additional_accrued_interest  hardship_payoff_balance_amount  \\\n",
       "0                                         NaN                             NaN   \n",
       "1                                         NaN                             NaN   \n",
       "2                                         NaN                             NaN   \n",
       "3                                         NaN                             NaN   \n",
       "4                                         NaN                             NaN   \n",
       "\n",
       "   hardship_last_payment_amount disbursement_method debt_settlement_flag  \\\n",
       "0                           NaN                Cash                    N   \n",
       "1                           NaN                Cash                    N   \n",
       "2                           NaN                Cash                    N   \n",
       "3                           NaN                Cash                    N   \n",
       "4                           NaN                Cash                    N   \n",
       "\n",
       "  debt_settlement_flag_date settlement_status settlement_date  \\\n",
       "0                       NaN               NaN             NaN   \n",
       "1                       NaN               NaN             NaN   \n",
       "2                       NaN               NaN             NaN   \n",
       "3                       NaN               NaN             NaN   \n",
       "4                       NaN               NaN             NaN   \n",
       "\n",
       "   settlement_amount  settlement_percentage  settlement_term  \n",
       "0                NaN                    NaN              NaN  \n",
       "1                NaN                    NaN              NaN  \n",
       "2                NaN                    NaN              NaN  \n",
       "3                NaN                    NaN              NaN  \n",
       "4                NaN                    NaN              NaN  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with pd.option_context('display.max_columns', 1000):\n",
    "    display(df.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert relevant date columns\n",
    "df[\"issue_d\"] = pd.to_datetime(df[\"issue_d\"], format=\"%b-%Y\")\n",
    "df[\"last_pymnt_d\"] = pd.to_datetime(df[\"last_pymnt_d\"], format=\"%b-%Y\")\n",
    "\n",
    "# convert term to numeric -- careful of the space before \"x months\"\n",
    "df[\"term\"] = df[\"term\"].map({\" 36 months\":36, \" 60 months\":60})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining the outcome (target variable)\n",
    "Our goal is to reduce financial risk by correctly identifying \"bad\" loans. We need to choose a definition for \"bad\" loans. In general, this could be \"whether a loan has defaulted before the end of its term\", \"whether a loan is ever more than 60 days delinquent\", etc. Different definitions may have their own merits; for example if early delinquency is very common then it may make sense to define \"bad\" as \"60 days delinquent within the first year of origination\" in order to reduce losses from these loans.\n",
    "\n",
    "Let's take a look at the `loan_status` column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Fully Paid                                             1076751\n",
       "Current                                                 878317\n",
       "Charged Off                                             268559\n",
       "Late (31-120 days)                                       21467\n",
       "In Grace Period                                           8436\n",
       "Late (16-30 days)                                         4349\n",
       "Does not meet the credit policy. Status:Fully Paid        1988\n",
       "Does not meet the credit policy. Status:Charged Off        761\n",
       "Default                                                     40\n",
       "Name: loan_status, dtype: int64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"loan_status\"].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are 9 different statuses, although \"Fully Paid\" and \"Charged Off\" seem to be duplicated (LendingClub does not explain what the \"Does not meet...\" options mean, and some suspect it's a data entry mistake https://forum.lendacademy.com/?topic=2427.msg20813#msg20813). \n",
    "\n",
    "Note: in our data that we do not have access to the full transaction history during the loan. Thus for loans which have completed (which is a majority of the data), we only know their final status. Therefore, we will filter out incomplete loans and not have to worry about `Current`, `Late`, `In Grace Period`, or `Default`. We will define `Charged Off` loans to be \"bad\" and `Fully Paid` as \"good\"."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# let's take a look at a subset of the columns\n",
    "df_sub = df[[\"issue_d\",\"term\",\"loan_status\",\"last_pymnt_d\",\"int_rate\",\"fico_range_high\",\"dti\"]]\n",
    "df_sub = df_sub.loc[~df_sub.isna().any(axis=1)] # remove rows with any null entries\n",
    "\n",
    "# this isn't the encoding we will use for model-building, just for getting a sense of the whole dataset over time\n",
    "bad = ['Charged Off','In Grace Period', 'Late (31-120 days)', 'Late (16-30 days)', 'Default', 'Does not meet the credit policy. Status:Charged Off']\n",
    "good = ['Fully Paid', 'Current','Does not meet the credit policy. Status:Fully Paid']\n",
    "m = {b:1 for b in bad} # build a map to convert bad loans to 1\n",
    "m.update({g:0 for g in good}) # convert good loans to 0\n",
    "df_sub[\"loan_status\"] = df_sub[\"loan_status\"].map(m)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Timestamp('2019-03-01 00:00:00')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_sub[\"last_pymnt_d\"].max() # data looks to be up to date as of 2019-03-01"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train/validation/test split -- the importance of time\n",
    "In a real-world instance of this problem (predicting loan defaults), we would obviously want to apply our model on future loans (referred to as \"out-of-time data\"). However, many people neglect this when doing personal projects and instead just randomly partition their data into training, validation, and testing sets (say with respective sizes 70%, 15%, 15% of the original data) and do not consider the temporal nature of the problem. \n",
    "\n",
    "Aside from this method being unpractical, it also leads to data leakage and thus over-optimistic model evaluation. The reason for this is that features (interest rate, credit score, debt-to-income ratio, etc.) are non-stationary among the population of borrowers (in fact, the target variable may also change in distribution). The plots below confirm this with our data. We group loans by origination month and compute averages for several features (interest rate, credit score, debt-to-income ratio, and loan status)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "stats_over_time = df_sub.set_index(\"issue_d\").groupby(pd.Grouper(freq=\"M\"))[[\"int_rate\",\"fico_range_high\",\"dti\",\"loan_status\"]].mean()\n",
    "fig, ax = plt.subplots(4,1,figsize=(10,6))\n",
    "for i,col in enumerate([\"int_rate\",\"fico_range_high\",\"dti\",\"loan_status\"]):\n",
    "    plt.sca(ax[i])\n",
    "    stats_over_time[col].plot()\n",
    "    plt.ylabel(col)\n",
    "    \n",
    "plt.tight_layout()\n",
    "plt.savefig(\"images/time_series.png\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We see all of the features change over time to some extent. The most notable pattern is that debt-to-income ratio has steadily increased from 10% to 20% over LendingClub's lifetime, indicating that the loans issued just before shutting down may be riskier. The decline in `loan_status` (which measures the fraction of loans which are bad) following 2016 is artificial -- rather it is a result of these loans not being complete. \n",
    "\n",
    "----\n",
    "\n",
    "Back to explaining why NOT doing a temporally disjoint train-test split will lead to over-optimistic evaluation ⤵️ \n",
    "\n",
    "Suppose we do use an out-of-time test set. For example, use 3-year loans originating in 2010 for training, 3-year loans originating in 2014 for testing, and aim to predict default by the end of the loan term. Suppose then we build a naive model that predicts \"default\" if a borrower's credit score is below some threshold (say the 25th percentile in 2010, which is around 600) and \"non-default\" if their credit score is above 600. Now let's try to evaluate our model on the test set -- but suppose the 25th percentile credit score has rose to 650, but the relative amount of defaults has not changed. Many loans are going to be incorrectly classified as \"non-default\", leading to poorer -- but realistic -- performance.\n",
    "\n",
    "Note: some practitioners try to account for time, but do it improperly. For example, I've seen people split their train, validation, and testing sets by origination date (say loans with dates up to the 70th percentile go to the training set, loans with dates between 70th and 85th percentile go to validation, and loans with dates greater than the 85th percentile go to testing). This fails to note the fact that some loans in the training set will not have finished before the first loan in the validation set, and similar for the validation and testing sets.\n",
    "\n",
    "**We want temporally disjoint training and testing sets**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before we define our training/validation/testing datasets, we should note that there are two different loan terms (3 and 5 years). Three year loans are much more common, representing 71% of the data (see cell below). Because of this and because our dataset only consists of loans from 2007-2019 which we think is insufficient for making such long forecasts<sup>1</sup>, **we choose to focus on three year loans**.\n",
    "\n",
    "&#8291;1. Constructing appopriate *out-of-time* training/validation/testing sets (read more about this in the next section) would require at least 15 years of data which we do not have (even if we used in-time validation, we would only be able to use loans originating in the first few months of the data to train on which would severely limit our data. Furthermore, making 5 year default predictions is incredibly hard -- generally it is better to improve upon short (1-2 year) forecasts rather than trying to make 5-10 year forecasts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "36.0    0.712138\n",
       "60.0    0.287862\n",
       "Name: term, dtype: float64"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_sub[\"term\"].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Three ways to split\n",
    "Including the \"improper way\", there are three different ways we could split the data. We provide visualizations of the time windows for each method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# a helper function to conver a datetime to a float\n",
    "# e.g. 2017-07-01 --> 2017.5\n",
    "def datetime2year(dt): \n",
    "    year_part = dt - datetime(year=dt.year, month=1, day=1)\n",
    "    year_length = (\n",
    "        datetime(year=dt.year + 1, month=1, day=1)\n",
    "        - datetime(year=dt.year, month=1, day=1)\n",
    "    )\n",
    "    return dt.year + year_part / year_length"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. In-time testing\n",
    "Randomly split data into training/validation/testing sets. By \"in-time\" testing, we mean that the model is evaluated on loans originating from the same time period as the training/validation data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def add_legend(plt):\n",
    "    handles, labels = plt.gca().get_legend_handles_labels()\n",
    "    p1 = mpatches.Patch(color='C0', label='loan lifetime')\n",
    "    p2 = mpatches.Patch(color='red', label='origination window')\n",
    "    handles += [p1,p2]\n",
    "    plt.legend(handles=handles, loc='upper left')\n",
    "    \n",
    "start, end = datetime2year(df_sub[\"issue_d\"].min()), datetime2year(df_sub[\"last_pymnt_d\"].max())\n",
    "plt.barh(y=[0], width=[end-start], left=[start], height=0.25)\n",
    "plt.barh(y=[0], width=[end-start-3], left=[start], height=0.125, color=\"red\")\n",
    "plt.yticks([0], [\"Train+Val+Test\"])\n",
    "plt.xlim([start-0.5,end+0.5])\n",
    "plt.ylim([-1,1])\n",
    "add_legend(plt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Out-of-time testing, in-time validation\n",
    "Testing is done on future loans, but training and validation sets are randomly sampled from all loans **prior**. By prior we mean all loans ending before the the first origination date in the testing set.\n",
    "\n",
    "Because we generally want more training data than validation/testing data, it makes sense to use a small origination window to specify loans in the out-of-time testing set. We choose to use one quarter (3 months) as the origination window for the testing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_end = end-3.25\n",
    "plt.barh(y=[0,1], width=[train_end-start,3.25], left=[start,train_end], height=0.25)\n",
    "plt.barh(y=[0,1], width=[train_end-start-3,.25], left=[start,train_end], height=0.125, color=\"red\")\n",
    "plt.yticks([0,1], [\"Train+Val\", \"Test\"])\n",
    "plt.xlim([start-1,end+1])\n",
    "plt.ylim([-1,2])\n",
    "add_legend(plt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Out-of-time testing, out-of-time validation\n",
    "Testing **and** validation are done on future loans."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "val_end = end-3.25\n",
    "train_end = end-3.25*2\n",
    "plt.barh(y=[0,1,2], width=[train_end-start,3.25,3.25], left=[start,val_end-3.25,val_end], height=0.25)\n",
    "plt.barh(y=[0,1,2], width=[train_end-start-3,.25,.25], left=[start,val_end-3.25,val_end], height=0.125, color=\"red\")\n",
    "plt.yticks([0,1,2], [\"Train\", \"Val\", \"Test\"])\n",
    "plt.xlim([start-1,end+1])\n",
    "plt.ylim([-1,3])\n",
    "add_legend(plt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will not use method 1 due to the reasons we discussed. When comparing method 2 and method 3, note that method 2 has the advantage of allowing for more training data, but the disadvantage is that our model may overfit to the training/validation data and perform poorly on the out-of-time testing data. We would hope that our model would learn to not rely heavily on features that change rapidly in time (i.e. we do not want a naive credit score cutoff model), but using in-time validation will not expose our model to unseen conditions. In some sense, method 3 may be able to regularize our model. At the same time, if the validation window is not like the training window, but the test window is similar to the training window (i.e. if credit score is higher in validation, but back down during testing) then this may harm the model. \n",
    "\n",
    "Another consideration is that LendingClub grew a lot over the time range of our dataset (see plot below). Very few loans were issued pre 2010 and, from an earlier plot showing loan status vs issue date, we saw that these loans were more likely to \"go bad\". This presents a major disadvantage of method 3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='issue_d'>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_sub.groupby(df_sub[\"issue_d\"].dt.year).count()[\"issue_d\"].plot(kind=\"bar\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll choose method 2 because it offers us much more data. That is, for our testing set we will choose loans originating in 2016Q1. For our training & validation sets we will then find all three-year loans ending before 2016Q1 (so the last valid origination date should be 2012-12-31). From this, we'll then randomly sample 80-90% as training and the rest as validation. \n",
    "\n",
    "Note: we could instead use k-fold cross-validation (with k > 2); however, because we have a decent amount of data and because we will be applying a considerable amount of preprocessing & feature selection (which should be done separately on each training fold), resulting in increased computation time -- we will stick to 2-fold validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of loans in testing set 89784\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Fully Paid     75326\n",
       "Charged Off    14458\n",
       "Name: loan_status, dtype: int64"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# testing: three-year loans originating in 2016Q1\n",
    "test = df.loc[(df[\"issue_d\"] >= datetime(2016,1,1)) & \\\n",
    "              (df[\"issue_d\"] < datetime(2016,4,1)) & \\\n",
    "              (df[\"term\"]==36) & \\\n",
    "              (df[\"loan_status\"].isin([\"Fully Paid\", \"Charged Off\"]))]\n",
    "\n",
    "print(\"Number of loans in testing set\", test.shape[0])\n",
    "test[\"loan_status\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of loans in training set 58052\n",
      "Number of loans in validation set 14514\n"
     ]
    }
   ],
   "source": [
    "# training & validation: three-year loans originating prior to 2013\n",
    "tr_val = df.loc[(df[\"issue_d\"] < datetime(2012,12,31)) & \\\n",
    "                (df[\"term\"]==36) & \\\n",
    "                (df[\"loan_status\"].isin([\"Fully Paid\", \"Charged Off\"]))]\n",
    "\n",
    "# random sample\n",
    "idx = np.arange(len(tr_val))\n",
    "np.random.shuffle(idx)\n",
    "TR_SIZE = int(0.8*len(tr_val))\n",
    "tr = tr_val.iloc[idx[:TR_SIZE],:]\n",
    "val = tr_val.iloc[idx[TR_SIZE:],:]\n",
    "\n",
    "print(\"Number of loans in training set\", tr.shape[0])\n",
    "print(\"Number of loans in validation set\", val.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our training/validation/testing sets combined contain around 150k loans, which is a lot but also a far smaller amount than the 2 million in the whole dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# write to csv\n",
    "tr.to_csv(\"train.csv\", index = False)\n",
    "val.to_csv(\"val.csv\", index = False)\n",
    "test.to_csv(\"test.csv\", index = False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}