FS_challenge_SaraPereira.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# James Challenge Solution \n",
    "## by Sara Pereira\n",
    "\n",
    "The data from Lending Club 2016 is read from .csv files and consolidated into a pandas dataframe. \n",
    "\n",
    "The proposed questions are answered using basics methods of pandas. \n",
    "When exploring data it is seen how is the volume distribution of loan amounts, and how the grade of loans affects the target variable.\n",
    "\n",
    "This data set has a very high dimensionality so it has to be reduced. Variables that are mostly missing data are removed. It is then analysed the correlation between numerical variables, and some more are eliminated. Dates are turned into time periods and categorical variables are encoded. \n",
    "\n",
    "The data set is split into training set and test set, and it is ensured that every scaler or model is only fitted to the training set, and then applied to the test set.\n",
    "\n",
    "After standardising the data, some models are compared for the prediction of the loan status. The choice of models was done so it would contain classification models of different types (implementing different techniques).\n",
    "The results are evaluated in terms of recall, precision, f1 score mathews correlation coefficient and AUC with a ROC. \n",
    "All models deliver extraordinary performance, resulting from careful feature engineering.\n",
    "\n",
    "Some comments and analysis are made along the script.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import Statements"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import math\n",
    "\n",
    "# time\n",
    "import datetime\n",
    "from monthdelta import *\n",
    "\n",
    "# visualization\n",
    "import seaborn as sns\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "#models\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "# these models take too long to process\n",
    "#from sklearn.tree import DecisionTreeClassifier \n",
    "#from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier\n",
    "\n",
    "#evaluation metrics\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn import metrics\n",
    "from sklearn import preprocessing\n",
    "%matplotlib inline "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Read and Merge Files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfuly read LoanStats/LoanStats_2016Q1.csv. Data corresponds to the following issuing months:\n",
      "Mar-2016    61992\n",
      "Feb-2016    39529\n",
      "Jan-2016    32366\n",
      "Name: issue_d, dtype: int64\n",
      "Successfuly read LoanStats/LoanStats_2016Q2.csv. Data corresponds to the following issuing months:\n",
      "Apr-2016    36432\n",
      "Jun-2016    33019\n",
      "May-2016    28403\n",
      "Name: issue_d, dtype: int64\n",
      "Successfuly read LoanStats/LoanStats_2016Q3.csv. Data corresponds to the following issuing months:\n",
      "Aug-2016    36280\n",
      "Jul-2016    34696\n",
      "Sep-2016    28144\n",
      "Name: issue_d, dtype: int64\n",
      "Successfuly read LoanStats/LoanStats_2016Q4.csv. Data corresponds to the following issuing months:\n",
      "Dec-2016    36183\n",
      "Nov-2016    34591\n",
      "Oct-2016    32772\n",
      "Name: issue_d, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "files = ['LoanStats/LoanStats_2016Q1.csv', 'LoanStats/LoanStats_2016Q2.csv', 'LoanStats/LoanStats_2016Q3.csv', 'LoanStats/LoanStats_2016Q4.csv']\n",
    "frames = ['Q1', 'Q2', 'Q3', 'Q4']\n",
    "ddf = {}\n",
    "\n",
    "for frame, file in zip(frames, files):\n",
    "    ddf[frame] = pd.read_csv(file, skiprows=1, low_memory=False)\n",
    "    print('Successfuly read {}. Data corresponds to the following issuing months:'.format(file))\n",
    "    print(ddf[frame]['issue_d'].value_counts())\n",
    "\n",
    "df = pd.concat(ddf)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></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>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 rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>12000.0</td>\n",
       "      <td>12000.0</td>\n",
       "      <td>12000.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>6.97%</td>\n",
       "      <td>237.45</td>\n",
       "      <td>A</td>\n",
       "      <td>A3</td>\n",
       "      <td>...</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>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>8000.0</td>\n",
       "      <td>8000.0</td>\n",
       "      <td>8000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>11.99%</td>\n",
       "      <td>265.68</td>\n",
       "      <td>C</td>\n",
       "      <td>C1</td>\n",
       "      <td>...</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>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>8200.0</td>\n",
       "      <td>8200.0</td>\n",
       "      <td>8200.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>11.47%</td>\n",
       "      <td>270.29</td>\n",
       "      <td>B</td>\n",
       "      <td>B5</td>\n",
       "      <td>...</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>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>14000.0</td>\n",
       "      <td>14000.0</td>\n",
       "      <td>14000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>10.75%</td>\n",
       "      <td>456.69</td>\n",
       "      <td>B</td>\n",
       "      <td>B4</td>\n",
       "      <td>...</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>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>15000.0</td>\n",
       "      <td>15000.0</td>\n",
       "      <td>15000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>9.16%</td>\n",
       "      <td>478.12</td>\n",
       "      <td>B</td>\n",
       "      <td>B2</td>\n",
       "      <td>...</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",
       "<p>5 rows × 145 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       id  member_id  loan_amnt  funded_amnt  funded_amnt_inv        term  \\\n",
       "Q1 0  NaN        NaN    12000.0      12000.0          12000.0   60 months   \n",
       "   1  NaN        NaN     8000.0       8000.0           8000.0   36 months   \n",
       "   2  NaN        NaN     8200.0       8200.0           8200.0   36 months   \n",
       "   3  NaN        NaN    14000.0      14000.0          14000.0   36 months   \n",
       "   4  NaN        NaN    15000.0      15000.0          15000.0   36 months   \n",
       "\n",
       "     int_rate  installment grade sub_grade       ...        \\\n",
       "Q1 0    6.97%       237.45     A        A3       ...         \n",
       "   1   11.99%       265.68     C        C1       ...         \n",
       "   2   11.47%       270.29     B        B5       ...         \n",
       "   3   10.75%       456.69     B        B4       ...         \n",
       "   4    9.16%       478.12     B        B2       ...         \n",
       "\n",
       "     hardship_payoff_balance_amount hardship_last_payment_amount  \\\n",
       "Q1 0                            NaN                          NaN   \n",
       "   1                            NaN                          NaN   \n",
       "   2                            NaN                          NaN   \n",
       "   3                            NaN                          NaN   \n",
       "   4                            NaN                          NaN   \n",
       "\n",
       "     disbursement_method  debt_settlement_flag debt_settlement_flag_date  \\\n",
       "Q1 0                Cash                     N                       NaN   \n",
       "   1                Cash                     N                       NaN   \n",
       "   2                Cash                     N                       NaN   \n",
       "   3                Cash                     N                       NaN   \n",
       "   4                Cash                     N                       NaN   \n",
       "\n",
       "     settlement_status settlement_date settlement_amount  \\\n",
       "Q1 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_percentage settlement_term  \n",
       "Q1 0                    NaN             NaN  \n",
       "   1                    NaN             NaN  \n",
       "   2                    NaN             NaN  \n",
       "   3                    NaN             NaN  \n",
       "   4                    NaN             NaN  \n",
       "\n",
       "[5 rows x 145 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Questions "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### a. How many loans are *there*?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "434415"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### b. How many columns does the dataset have?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "145"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### c. How many columns only contain missing data?\n",
    "\n",
    "Columns only containing missing data won't be of help so we might as well just eliminate them "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 145 - 132 = 13\n"
     ]
    }
   ],
   "source": [
    "df.dropna(axis=1, how='all', inplace=True) \n",
    "print('Removed column(s) = 145 - {} = {}'.format(df.shape[1], (145-df.shape[1]))) # 145 is the original number of columns "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### d. How many loans are still current? \n",
    "\n",
    "'Current' is one of the possible loan status with 225496 instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Current               225496\n",
       "Fully Paid            145318\n",
       "Charged Off            50819\n",
       "Late (31-120 days)      6728\n",
       "In Grace Period         4518\n",
       "Late (16-30 days)       1517\n",
       "Default                   11\n",
       "Name: loan_status, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['loan_status'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Subset Exploration\n",
    "\n",
    "The exploration takes into account that the goal is to predict the probability of a loan being charged off. \n",
    "\n",
    "From now on the variable 'loan_status' is considered the target variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Fully Paid     145318\n",
       "Charged Off     50819\n",
       "Name: loan_status, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# subset creation \n",
    "\n",
    "df = df[((df['loan_status'] == 'Fully Paid') | (df['loan_status'] == 'Charged Off'))] # the other categories won't be needed \n",
    "                                                                                      # so I'll just rewrite the df variable\n",
    "df['loan_status'].value_counts() # Charged Off status acounts for ~25% of the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "MultiIndex: 196137 entries, (Q1, 2) to (Q4, 103544)\n",
      "Columns: 132 entries, id to settlement_term\n",
      "dtypes: float64(94), object(38)\n",
      "memory usage: 199.5+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 132 - 130 = 2\n"
     ]
    }
   ],
   "source": [
    "# Remove columns with only missing data that may have surged after subsetting\n",
    "df.dropna(axis=1, how='all', inplace=True) \n",
    "print('Removed column(s) = 132 - {} = {}'.format(df.shape[1], (132-df.shape[1])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Distribution of Loan amount \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "50% of the loan amounts id under 12000.0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.loan_amnt.plot(kind=\"hist\", bins=15)\n",
    "print('50% of the loan amounts id under {}'.format(df.loan_amnt.median()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Loan grade volumes (filtered by loan status)\n",
    "The results suggest the loan grade may play an important role on the loan being charged off. As one can see there is a big difference between both distributions."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Charged Off Loans"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1fd809ba160>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df[df['loan_status'] == 'Charged Off'].grade.value_counts().plot(kind='bar', color = \n",
    "                                                                 ['black', 'red', 'green', 'blue', 'cyan', 'yellow', 'orange'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Fully paid loans"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1fd80b6d668>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df[df['loan_status'] == 'Fully Paid'].grade.value_counts().plot(kind='bar', color = \n",
    "                                                                ['green', 'black', 'yellow', 'red', 'blue', 'cyan', 'orange'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Variable exploration"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Variables that are mostly NaN \n",
    "The variables with only missing values were already removed. However, given the high dimensionality of the problem it is wise to remove the ones where more than half values are NaN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "desc                                          0.999924\n",
      "orig_projected_additional_accrued_interest    0.996992\n",
      "deferral_term                                 0.993601\n",
      "hardship_amount                               0.993601\n",
      "hardship_start_date                           0.993601\n",
      "hardship_end_date                             0.993601\n",
      "payment_plan_start_date                       0.993601\n",
      "hardship_length                               0.993601\n",
      "hardship_dpd                                  0.993601\n",
      "hardship_loan_status                          0.993601\n",
      "hardship_payoff_balance_amount                0.993601\n",
      "hardship_status                               0.993601\n",
      "hardship_last_payment_amount                  0.993601\n",
      "hardship_type                                 0.993601\n",
      "hardship_reason                               0.993601\n",
      "annual_inc_joint                              0.982543\n",
      "dti_joint                                     0.982543\n",
      "verification_status_joint                     0.982543\n",
      "settlement_percentage                         0.970857\n",
      "settlement_term                               0.970857\n",
      "debt_settlement_flag_date                     0.970857\n",
      "settlement_status                             0.970857\n",
      "settlement_date                               0.970857\n",
      "settlement_amount                             0.970857\n",
      "mths_since_last_record                        0.793563\n",
      "mths_since_recent_bc_dlq                      0.745377\n",
      "mths_since_last_major_derog                   0.704146\n",
      "mths_since_recent_revol_delinq                0.640919\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "#determine NaN percentage\n",
    "check_null = df.isnull().sum().sort_values(ascending=False)/len(df)\n",
    "\n",
    "#print all with more than 50% NaNs\n",
    "print(check_null[check_null > 0.50])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 130 - 102 = 28\n"
     ]
    }
   ],
   "source": [
    "# Remove more than 50% NaN variables\n",
    "df.drop(check_null[check_null > 0.5].index, axis=1, inplace=True)\n",
    "df.dropna(axis=0, thresh=30,inplace=True)\n",
    "print('Removed column(s) = 130 - {} = {}'.format(df.shape[1], (130-df.shape[1])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Variable Correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x648 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Select numerical variables\n",
    "num = df.select_dtypes(include=np.number)\n",
    "names = num.columns.values\n",
    "\n",
    "# Compute the correlation matrix\n",
    "corr = num.corr()\n",
    "\n",
    "# Visualization\n",
    "sns.set(style=\"white\")\n",
    "# Generate a mask for the upper triangle\n",
    "mask = np.zeros_like(corr, dtype=np.bool)\n",
    "mask[np.triu_indices_from(mask)] = True\n",
    "\n",
    "# Set up the matplotlib figure\n",
    "f, ax = plt.subplots(figsize=(11, 9))\n",
    "\n",
    "# Generate a custom diverging colormap\n",
    "cmap = 'Spectral_r'\n",
    "# sns.diverging_palette(100, 275 , as_cmap=True)\n",
    "\n",
    "# Draw the heatmap with the mask and correct aspect ratio\n",
    "sns.heatmap(corr, mask=mask, cmap=cmap, vmax=1, center=0,\n",
    "            square=True, linewidths=.5, cbar_kws={\"shrink\": .5})\n",
    "\n",
    "#ax.tick_params(labelsize=8)\n",
    "plt.yticks(range(num.shape[1]), names,rotation=0, size=6)\n",
    "#plt.xticks(rotation=90)\n",
    "plt.xticks(range(num.shape[1]), names, rotation=90, size=6)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " Correlation is not causation. However, since the number of numerical variables is so high we can still check the pairs where the absolute correlation value is higher than 85% and discard some high correlated variables on the basis that they would offer approximately the same information, being redundant. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              level_0                     level_1         0\n",
      "0           loan_amnt                 funded_amnt  1.000000\n",
      "1           loan_amnt             funded_amnt_inv  0.999992\n",
      "2           loan_amnt                 installment  0.956304\n",
      "3         funded_amnt             funded_amnt_inv  0.999992\n",
      "4         funded_amnt                 installment  0.956304\n",
      "5     funded_amnt_inv                 installment  0.956220\n",
      "6            open_acc                    num_sats  0.998653\n",
      "7         total_pymnt             total_pymnt_inv  0.999994\n",
      "8         total_pymnt             total_rec_prncp  0.979293\n",
      "9     total_pymnt_inv             total_rec_prncp  0.979288\n",
      "10         recoveries     collection_recovery_fee  0.987253\n",
      "11        tot_cur_bal             tot_hi_cred_lim  0.957448\n",
      "12       total_bal_il           total_bal_ex_mort  0.895452\n",
      "13       total_bal_il  total_il_high_credit_limit  0.960351\n",
      "14    num_actv_rev_tl         num_rev_tl_bal_gt_0  0.975562\n",
      "15  total_bal_ex_mort  total_il_high_credit_limit  0.881402\n"
     ]
    }
   ],
   "source": [
    "dataCorr = corr.mask(np.tril(np.ones(corr.shape)).astype(np.bool))\n",
    "dataCorr = dataCorr[abs(dataCorr) >= 0.85].stack().reset_index()\n",
    "print(dataCorr[:50])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- The correlation between 'loan_amnt', 'funded_amnt', and 'funded_amnt_inv' is avery high, so let's keep only the 'funded_amnt' which represents the amount actually funded\n",
    "- let's also remove  'loan_amnt', 'installment', 'num_sats', 'total_pymnt_inv', 'total_rec_prncp', 'collection_recovery_fee', 'tot_hi_cred_lim', 'total_bal_ex_mort', 'total_il_high_credit_limit' and 'num_rev_tl_bal_gt_0'. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 102 - 92 = 10\n"
     ]
    }
   ],
   "source": [
    "df.drop(['loan_amnt', 'installment', 'num_sats', 'total_pymnt_inv', 'total_rec_prncp', 'collection_recovery_fee', \n",
    "         'tot_hi_cred_lim', 'total_bal_ex_mort', 'total_il_high_credit_limit', 'num_rev_tl_bal_gt_0'], axis = 1, inplace = True)\n",
    "\n",
    "print('Removed column(s) = 102 - {} = {}'.format(df.shape[1], (102-df.shape[1])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By now the variables were reduced <b>from 145 to 92</b> by removing high NaN percentage variables, and highly correlated. \n",
    "Let's check the remaining variables for categorical data types and for more reduction possibilities. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Column 'term' has 2 unique instances ([' 36 months' ' 60 months']) . \n",
      "\n",
      "Column 'int_rate' has 138 unique instances ([' 11.47%' '  9.16%' ' 14.46%' ' 10.75%' ' 13.67%' ' 18.25%' ' 15.31%'\n",
      " ' 11.99%' ' 19.53%' ' 16.29%' '  5.32%' ' 12.99%' ' 24.11%' '  7.39%'\n",
      " ' 17.27%' '  9.75%' '  8.39%' ' 22.45%' ' 21.18%' '  7.89%' ' 19.99%'\n",
      " '  6.49%' ' 20.75%' ' 18.99%' '  6.97%' ' 28.34%' ' 21.97%' ' 24.99%'\n",
      " ' 26.57%' ' 28.99%' ' 28.14%' '  6.00%' ' 23.13%' ' 28.67%' ' 25.88%'\n",
      " ' 27.34%' ' 15.29%' ' 10.78%' '  9.17%' ' 12.88%' ' 13.44%' '  7.91%'\n",
      " ' 11.48%' ' 13.99%' '  6.99%' '  8.49%' ' 24.24%' ' 26.99%' ' 16.59%'\n",
      " ' 18.49%']) . Only showing first 50. \n",
      "\n",
      "Column 'grade' has 7 unique instances (['B' 'C' 'D' 'A' 'F' 'E' 'G']) . \n",
      "\n",
      "Column 'sub_grade' has 35 unique instances (['B5' 'B2' 'C4' 'B4' 'C3' 'D3' 'C5' 'C1' 'D5' 'D1' 'A1' 'C2' 'F2' 'A4'\n",
      " 'D2' 'B3' 'B1' 'E5' 'E3' 'A5' 'E1' 'A2' 'E2' 'D4' 'A3' 'G3' 'E4' 'F3'\n",
      " 'F5' 'G5' 'G2' 'F1' 'G4' 'F4' 'G1']) . \n",
      "\n",
      "Column 'emp_title' has 64407 unique instances (['Registered Nurse' 'mainteance' 'Fire fighter' nan 'parts sales'\n",
      " 'Senior Client Associate' 'Facility Scheduling & EVents Manager'\n",
      " 'vice president' 'Store Manager' 'Wealth Manager' 'Patrolman'\n",
      " 'Vice principal' 'Nurse' 'ASSISTANT CONTROLLER' 'DEPARTMENT MANAGER'\n",
      " 'Attorney ' 'Direct Support Staff' 'member' 'Receiving Operator'\n",
      " 'Controls Engineer ' 'College Assistant'\n",
      " 'coordinator Materials Management' 'Armed guard' 'Scheduler'\n",
      " 'facility administrator' 'Assembly' 'Crane Operator ' 'Project Manager'\n",
      " 'Driver' 'Trans Mgmt Coordinator 88N' 'General manager' 'Teacher'\n",
      " 'Branch Manager' 'cook' 'Owner' 'driver' 'Kindergarten Teacher'\n",
      " 'Business Manager' 'Manager' 'installer' 'Customer service associate '\n",
      " 'Bookkeeper' 'supervisor ' 'Executive Chef' 'Parts Consultant'\n",
      " 'Training specialist ' 'Table Games Supervisor' 'Officer'\n",
      " 'Senior Technical Support Engineer' 'Master Optician ']) . Only showing first 50. \n",
      "\n",
      "Column 'emp_length' has 12 unique instances (['10+ years' '2 years' nan '7 years' '9 years' '< 1 year' '4 years'\n",
      " '5 years' '1 year' '3 years' '6 years' '8 years']) . \n",
      "\n",
      "Column 'home_ownership' has 4 unique instances (['RENT' 'MORTGAGE' 'OWN' 'ANY']) . \n",
      "\n",
      "Column 'verification_status' has 3 unique instances (['Source Verified' 'Not Verified' 'Verified']) . \n",
      "\n",
      "Column 'issue_d' has 12 unique instances (['Mar-2016' 'Feb-2016' 'Jan-2016' 'Jun-2016' 'May-2016' 'Apr-2016'\n",
      " 'Sep-2016' 'Aug-2016' 'Jul-2016' 'Dec-2016' 'Nov-2016' 'Oct-2016']) . \n",
      "\n",
      "Column 'loan_status' has 2 unique instances (['Fully Paid' 'Charged Off']) . \n",
      "\n",
      "Column 'pymnt_plan' has 1 unique instances (['n']) . \n",
      "\n",
      "Column 'purpose' has 13 unique instances (['debt_consolidation' 'home_improvement' 'other' 'credit_card' 'moving'\n",
      " 'major_purchase' 'house' 'vacation' 'medical' 'small_business' 'car'\n",
      " 'renewable_energy' 'wedding']) . \n",
      "\n",
      "Column 'title' has 13 unique instances (['Debt consolidation' 'Home improvement' 'Other' 'Credit card refinancing'\n",
      " 'Moving and relocation' 'Major purchase' nan 'Vacation'\n",
      " 'Medical expenses' 'Business' 'Car financing' 'Home buying' 'Green loan']) . \n",
      "\n",
      "Column 'zip_code' has 896 unique instances (['391xx' '285xx' '799xx' '315xx' '281xx' '112xx' '820xx' '606xx' '116xx'\n",
      " '920xx' '463xx' '330xx' '328xx' '913xx' '957xx' '287xx' '200xx' '029xx'\n",
      " '815xx' '450xx' '978xx' '775xx' '104xx' '787xx' '950xx' '453xx' '335xx'\n",
      " '481xx' '234xx' '651xx' '970xx' '296xx' '750xx' '339xx' '609xx' '320xx'\n",
      " '440xx' '325xx' '770xx' '760xx' '871xx' '922xx' '908xx' '989xx' '917xx'\n",
      " '891xx' '380xx' '492xx' '186xx' '333xx']) . Only showing first 50. \n",
      "\n",
      "Column 'addr_state' has 50 unique instances (['MS' 'NC' 'TX' 'GA' 'NY' 'WY' 'IL' 'CA' 'IN' 'FL' 'DC' 'RI' 'CO' 'OH'\n",
      " 'OR' 'MI' 'VA' 'MO' 'SC' 'NM' 'WA' 'NV' 'TN' 'PA' 'KS' 'WV' 'AZ' 'MN'\n",
      " 'MD' 'MT' 'MA' 'CT' 'ND' 'WI' 'AR' 'NJ' 'KY' 'VT' 'UT' 'SD' 'AL' 'NH'\n",
      " 'OK' 'ID' 'LA' 'AK' 'NE' 'HI' 'DE' 'ME']) . \n",
      "\n",
      "Column 'earliest_cr_line' has 645 unique instances (['Jun-2006' 'Dec-2003' 'May-2005' 'Oct-1989' 'Nov-1983' 'Oct-1997'\n",
      " 'Jun-2005' 'Mar-1983' 'Sep-2006' 'Apr-1995' 'Jan-2005' 'Jan-1992'\n",
      " 'Sep-2008' 'Jul-2001' 'Nov-1987' 'Apr-2001' 'Aug-2004' 'Aug-2012'\n",
      " 'Dec-2000' 'Mar-1985' 'Mar-2005' 'May-2001' 'Sep-2012' 'Aug-2003'\n",
      " 'Apr-2007' 'Jul-1999' 'Sep-1991' 'Aug-2001' 'Apr-2004' 'Mar-1996'\n",
      " 'Oct-2000' 'Dec-1999' 'Nov-1999' 'Feb-2006' 'Sep-2007' 'Jan-2002'\n",
      " 'Apr-1983' 'Jul-2008' 'Jun-1993' 'Aug-2006' 'Nov-2005' 'Jun-2011'\n",
      " 'Jun-1997' 'Apr-1998' 'Feb-2000' 'Nov-1995' 'Apr-2002' 'Nov-1991'\n",
      " 'Jul-2002' 'Jan-2001']) . Only showing first 50. \n",
      "\n",
      "Column 'revol_util' has 1118 unique instances (['41.7%' '58.6%' '94.6%' '48.9%' '73.9%' '46.3%' '54.2%' '77.4%' '40.7%'\n",
      " '31.5%' '16.3%' '76.2%' '7.2%' '45.9%' '84.4%' '97%' '16.1%' '9.7%'\n",
      " '56.7%' '53.1%' '12.6%' '61%' '12%' '3.4%' '66.3%' '82.2%' '63.5%'\n",
      " '55.2%' '34.2%' '101.6%' '80%' '53.2%' '19.9%' '10.7%' '84.8%' '48.5%'\n",
      " '81.4%' '62%' '85.9%' '32.9%' '35.6%' '31.8%' '58%' '59.8%' '69.3%'\n",
      " '52.9%' '62.4%' '48%' '87.3%' '79%']) . Only showing first 50. \n",
      "\n",
      "Column 'initial_list_status' has 2 unique instances (['w' 'f']) . \n",
      "\n",
      "Column 'last_pymnt_d' has 31 unique instances (['Dec-2017' 'Jan-2018' 'Sep-2016' 'Aug-2016' 'Feb-2017' 'Feb-2018'\n",
      " 'Aug-2017' 'Jul-2017' 'Mar-2017' 'Nov-2017' 'Jun-2017' 'Jun-2018'\n",
      " 'Oct-2017' 'Apr-2017' 'Sep-2017' 'Dec-2016' 'Oct-2016' 'Mar-2018'\n",
      " 'May-2017' 'Jul-2016' 'Nov-2016' 'Apr-2016' 'Jun-2016' 'May-2018'\n",
      " 'Apr-2018' 'Jan-2017' 'May-2016' nan 'Mar-2016' 'Feb-2016' 'Jan-2016']) . \n",
      "\n",
      "Column 'last_credit_pull_d' has 33 unique instances (['Apr-2018' 'Dec-2017' 'Dec-2016' 'Jun-2018' 'Aug-2017' 'Jun-2017'\n",
      " 'Nov-2017' 'Oct-2017' 'Jul-2017' 'Mar-2018' 'Apr-2017' 'Jan-2018'\n",
      " 'May-2017' 'Mar-2017' 'Sep-2017' 'May-2018' 'Feb-2017' 'Oct-2016'\n",
      " 'Feb-2018' 'Apr-2016' 'Jul-2016' 'Nov-2016' 'Jun-2016' 'May-2016'\n",
      " 'Aug-2016' 'Jan-2017' 'Sep-2016' 'Mar-2016' nan 'Feb-2016' 'Jan-2016'\n",
      " 'Dec-2015' 'Jul-2018']) . \n",
      "\n",
      "Column 'application_type' has 2 unique instances (['Individual' 'Joint App']) . \n",
      "\n",
      "Column 'hardship_flag' has 1 unique instances (['N']) . \n",
      "\n",
      "Column 'disbursement_method' has 2 unique instances (['Cash' 'DirectPay']) . \n",
      "\n",
      "Column 'debt_settlement_flag' has 2 unique instances (['N' 'Y']) . \n",
      "\n"
     ]
    }
   ],
   "source": [
    "for col in df.select_dtypes(include=['object']).columns:\n",
    "    print (\"Column '{}' has {} unique instances ({}) {} \\n\".format(\n",
    "        col, len(df[col].unique()), df[col].unique()[:50], \n",
    "        '.' if len(df[col].unique()) <= 50 else ('. Only showing first 50.')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- 'term'\n",
    "        -> TURN TO INT TYPE\n",
    "- 'int_rate' \n",
    "        -> TURN TO FLOAT TYPE\n",
    "- 'grade' \n",
    "        -> CATEGORICAL\n",
    "- 'subgrade'\n",
    "        -> CATEGORICAL\n",
    "- 'emp_title' free text, high cardinality.\n",
    "        -> REMOVE\n",
    "- 'emp_length' \n",
    "        -> TURN TO INT TYPE\n",
    "- 'home_ownership'\n",
    "        -> CATEGORICAL\n",
    "- 'verification_status' \n",
    "        -> CATEGORICAL\n",
    "- 'issue_d' \n",
    "        -> ELAPSED DATE       \n",
    "- 'loan_status' \n",
    "        -> REMOVE !!! TARGET VARIABLE !!! \n",
    "- 'pymnt_plan' constant column \n",
    "        -> REMOVE\n",
    "- 'purpose' \n",
    "        -> CATEGORICAL \n",
    "- 'title' same info as purpose\n",
    "        -> REMOVE\n",
    "- 'zip_code' categorical high cardinality \n",
    "        -> REMOVE\n",
    "- 'addr_state' categorical high cardinality\n",
    "        -> REMOVE\n",
    "- 'earliest_cr_line' \n",
    "        -> ELAPSED DATE\n",
    "- 'revol_util' \n",
    "        -> TURN TO FLOAT TYPE\n",
    "- 'initial_list_status' \n",
    "        -> CATEGORICAL\n",
    "- 'last_pymnt_d' not considered relevant\n",
    "        -> REMOVE \n",
    "- 'last_credit_pull_d' not considered relevant\n",
    "        -> REMOVE\n",
    "- 'application_type' \n",
    "        -> CATEGORICAL \n",
    "- 'hardship_flag' constant  \n",
    "        -> REMOVE\n",
    "- 'disbursement_method' \n",
    "        -> CATEGORICAL\n",
    "- 'debt_settlement_flag' \n",
    "        -> CATEGORICAL"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Variables 1 - 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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></th>\n",
       "      <th>funded_amnt</th>\n",
       "      <th>funded_amnt_inv</th>\n",
       "      <th>term</th>\n",
       "      <th>int_rate</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>...</th>\n",
       "      <th>application_type</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",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>2</th>\n",
       "      <td>8200.0</td>\n",
       "      <td>8200.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>11.47%</td>\n",
       "      <td>B</td>\n",
       "      <td>B5</td>\n",
       "      <td>Registered Nurse</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>RENT</td>\n",
       "      <td>70000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>Individual</td>\n",
       "      <td>0.0</td>\n",
       "      <td>149.0</td>\n",
       "      <td>37329.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>30992.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>15000.0</td>\n",
       "      <td>15000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>9.16%</td>\n",
       "      <td>B</td>\n",
       "      <td>B2</td>\n",
       "      <td>mainteance</td>\n",
       "      <td>2 years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>55000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>Individual</td>\n",
       "      <td>0.0</td>\n",
       "      <td>403.0</td>\n",
       "      <td>96061.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>12499.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>16000.0</td>\n",
       "      <td>16000.0</td>\n",
       "      <td>60 months</td>\n",
       "      <td>14.46%</td>\n",
       "      <td>C</td>\n",
       "      <td>C4</td>\n",
       "      <td>Fire fighter</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>69000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>Individual</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>205141.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>63369.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>8400.0</td>\n",
       "      <td>8400.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>10.75%</td>\n",
       "      <td>B</td>\n",
       "      <td>B4</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>OWN</td>\n",
       "      <td>24000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>Individual</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>23541.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>73.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>10000.0</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>36 months</td>\n",
       "      <td>11.47%</td>\n",
       "      <td>B</td>\n",
       "      <td>B5</td>\n",
       "      <td>parts sales</td>\n",
       "      <td>10+ years</td>\n",
       "      <td>MORTGAGE</td>\n",
       "      <td>55000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>Individual</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>46363.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>13286.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 50 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       funded_amnt  funded_amnt_inv        term int_rate grade sub_grade  \\\n",
       "Q1 2        8200.0           8200.0   36 months   11.47%     B        B5   \n",
       "   4       15000.0          15000.0   36 months    9.16%     B        B2   \n",
       "   8       16000.0          16000.0   60 months   14.46%     C        C4   \n",
       "   13       8400.0           8400.0   36 months   10.75%     B        B4   \n",
       "   17      10000.0          10000.0   36 months   11.47%     B        B5   \n",
       "\n",
       "              emp_title emp_length home_ownership  annual_inc      ...       \\\n",
       "Q1 2   Registered Nurse  10+ years           RENT     70000.0      ...        \n",
       "   4         mainteance    2 years       MORTGAGE     55000.0      ...        \n",
       "   8       Fire fighter  10+ years       MORTGAGE     69000.0      ...        \n",
       "   13               NaN        NaN            OWN     24000.0      ...        \n",
       "   17       parts sales  10+ years       MORTGAGE     55000.0      ...        \n",
       "\n",
       "      application_type acc_now_delinq tot_coll_amt tot_cur_bal open_acc_6m  \\\n",
       "Q1 2        Individual            0.0        149.0     37329.0         1.0   \n",
       "   4        Individual            0.0        403.0     96061.0         0.0   \n",
       "   8        Individual            0.0          0.0    205141.0         0.0   \n",
       "   13       Individual            0.0          0.0     23541.0         1.0   \n",
       "   17       Individual            0.0          0.0     46363.0         3.0   \n",
       "\n",
       "      open_act_il open_il_12m open_il_24m  mths_since_rcnt_il  total_bal_il  \n",
       "Q1 2          1.0         0.0         1.0                15.0       30992.0  \n",
       "   4          1.0         0.0         0.0                28.0       12499.0  \n",
       "   8          2.0         0.0         1.0                24.0       63369.0  \n",
       "   13         0.0         0.0         0.0                73.0           0.0  \n",
       "   17         1.0         1.0         2.0                 6.0       13286.0  \n",
       "\n",
       "[5 rows x 50 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.iloc[:5,:50]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# turn 'term' to int\n",
    "df.term = pd.Series(df.term).str.replace('months', '').astype(int)\n",
    "\n",
    "# turn 'int_rate' to float \n",
    "\n",
    "df.int_rate = pd.Series(df.int_rate).str.replace('%', '').astype(float)\n",
    "df.int_rate.values # array([11.47,  9.16, 14.46, ..., 14.49, 12.79, 14.49])\n",
    "\n",
    "# turn 'emp_length' to int using regex (NaN -> 0)\n",
    "\n",
    "df.emp_length.fillna(value=0,inplace=True)\n",
    "df.emp_length.replace(to_replace='[^0-9]+', value='', inplace=True, regex=True)\n",
    "df.emp_length = df.emp_length.astype(int)\n",
    "df.emp_length.dtype # dtype('int32')\n",
    "\n",
    "# turn 'issue_d', 'earliest_cr_line' to months elapsed since (...) -> see Date Variables\n",
    "\n",
    "# turn 'revol_util' to float \n",
    "\n",
    "df.revol_util = pd.Series(df.revol_util).str.replace('%', '').astype(float)\n",
    "\n",
    "# turn 'application_type' to binary\n",
    "df.application_type = pd.Series(df.application_type).str.replace('Individual', '1')\n",
    "df.application_type = pd.Series(df.application_type).str.replace('Joint App', '0').astype(int)\n",
    "\n",
    "# turn 'initial_list_status' to binary\n",
    "df.initial_list_status = pd.Series(df.initial_list_status).str.replace('w', '1')\n",
    "df.initial_list_status = pd.Series(df.initial_list_status).str.replace('f', '0').astype(int)\n",
    "\n",
    "# turn 'disbursement_method' to binary \n",
    "df.disbursement_method = pd.Series(df.disbursement_method).str.replace('Cash', '1')\n",
    "df.disbursement_method = pd.Series(df.disbursement_method).str.replace('DirectPay', '0').astype(int)\n",
    "\n",
    "# turn 'debt_settlement_flag' to binary \n",
    "df.debt_settlement_flag = pd.Series(df.debt_settlement_flag).str.replace('Y', '1')\n",
    "df.debt_settlement_flag = pd.Series(df.debt_settlement_flag).str.replace('N', '0').astype(int)\n",
    "\n",
    "# remove 'out_prncp' and 'out_prncp_inv' null variables\n",
    "\n",
    "# remove 'policy_code' constant column\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Variables 50 - 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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></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>...</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>total_bc_limit</th>\n",
       "      <th>hardship_flag</th>\n",
       "      <th>disbursement_method</th>\n",
       "      <th>debt_settlement_flag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>2</th>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>4536.0</td>\n",
       "      <td>42.0</td>\n",
       "      <td>15200.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>94.4</td>\n",
       "      <td>33.3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>9300.0</td>\n",
       "      <td>N</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>58.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1287.0</td>\n",
       "      <td>58.0</td>\n",
       "      <td>9600.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1400.0</td>\n",
       "      <td>N</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>79.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2187.0</td>\n",
       "      <td>82.0</td>\n",
       "      <td>7300.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>90.9</td>\n",
       "      <td>100.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2300.0</td>\n",
       "      <td>N</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>NaN</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>7085.0</td>\n",
       "      <td>49.0</td>\n",
       "      <td>48100.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>92.3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>33800.0</td>\n",
       "      <td>N</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>95.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2150.0</td>\n",
       "      <td>88.0</td>\n",
       "      <td>6000.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>68.7</td>\n",
       "      <td>75.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6000.0</td>\n",
       "      <td>N</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 42 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       il_util  open_rv_12m  open_rv_24m  max_bal_bc  all_util  \\\n",
       "Q1 2       NaN          1.0          3.0      4536.0      42.0   \n",
       "   4      58.0          0.0          2.0      1287.0      58.0   \n",
       "   8      79.0          0.0          1.0      2187.0      82.0   \n",
       "   13      NaN          1.0          3.0      7085.0      49.0   \n",
       "   17     95.0          2.0          3.0      2150.0      88.0   \n",
       "\n",
       "       total_rev_hi_lim  inq_fi  total_cu_tl  inq_last_12m  \\\n",
       "Q1 2            15200.0     1.0          0.0           1.0   \n",
       "   4             9600.0     0.0          0.0           0.0   \n",
       "   8             7300.0     0.0          6.0           1.0   \n",
       "   13           48100.0     0.0          0.0           1.0   \n",
       "   17            6000.0     0.0          0.0           2.0   \n",
       "\n",
       "       acc_open_past_24mths          ...           num_tl_90g_dpd_24m  \\\n",
       "Q1 2                    4.0          ...                          0.0   \n",
       "   4                    2.0          ...                          0.0   \n",
       "   8                    2.0          ...                          0.0   \n",
       "   13                   3.0          ...                          0.0   \n",
       "   17                   5.0          ...                          0.0   \n",
       "\n",
       "       num_tl_op_past_12m  pct_tl_nvr_dlq  percent_bc_gt_75  \\\n",
       "Q1 2                  1.0            94.4              33.3   \n",
       "   4                  0.0           100.0             100.0   \n",
       "   8                  0.0            90.9             100.0   \n",
       "   13                 1.0            92.3               0.0   \n",
       "   17                 3.0            68.7              75.0   \n",
       "\n",
       "       pub_rec_bankruptcies  tax_liens  total_bc_limit  hardship_flag  \\\n",
       "Q1 2                    0.0        5.0          9300.0              N   \n",
       "   4                    0.0        0.0          1400.0              N   \n",
       "   8                    0.0        0.0          2300.0              N   \n",
       "   13                   0.0        0.0         33800.0              N   \n",
       "   17                   0.0        0.0          6000.0              N   \n",
       "\n",
       "       disbursement_method  debt_settlement_flag  \n",
       "Q1 2                     1                     0  \n",
       "   4                     1                     0  \n",
       "   8                     1                     0  \n",
       "   13                    1                     0  \n",
       "   17                    1                     0  \n",
       "\n",
       "[5 rows x 42 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.iloc[:5,50:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# turn 'hardship_start_date', 'hardship_end_date', 'payment_plan_start_date', \n",
    "                            #'debt_settlement_flag_date', 'settlement_date' to periods -> see Date Variables\n",
    "\n",
    "# 'hardship_flag'-> Remove (constant)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pre -Processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### High Cardinality variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 92 - 81 = 10\n"
     ]
    }
   ],
   "source": [
    "# Remove high cardinality variables \n",
    "\n",
    "df.drop(['emp_title', 'pymnt_plan', 'title', 'zip_code', 'addr_state', 'out_prncp', 'out_prncp_inv', \n",
    "        'policy_code', 'hardship_flag', 'last_pymnt_d', 'last_credit_pull_d'], axis = 1, inplace = True)\n",
    "\n",
    "print('Removed column(s) = 92 - {} = {}'.format(df.shape[1], (91-df.shape[1])))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Time Variables\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(196137, 2)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dates = ['issue_d' , 'earliest_cr_line']\n",
    "dfdates = df[dates].copy()\n",
    "dfdates[dates] = '1-' + df[dates].astype(str) \n",
    "dfdates.head()\n",
    "dfdates.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "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></th>\n",
       "      <th>issue_d</th>\n",
       "      <th>earliest_cr_line</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>2</th>\n",
       "      <td>1-Mar-2016</td>\n",
       "      <td>1-Jun-2006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1-Mar-2016</td>\n",
       "      <td>1-Dec-2003</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1-Mar-2016</td>\n",
       "      <td>1-May-2005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1-Mar-2016</td>\n",
       "      <td>1-Oct-1989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>1-Mar-2016</td>\n",
       "      <td>1-Nov-1983</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          issue_d earliest_cr_line\n",
       "Q1 2   1-Mar-2016       1-Jun-2006\n",
       "   4   1-Mar-2016       1-Dec-2003\n",
       "   8   1-Mar-2016       1-May-2005\n",
       "   13  1-Mar-2016       1-Oct-1989\n",
       "   17  1-Mar-2016       1-Nov-1983"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dfdates.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Turn columns to datetime objects\n",
    "dfdates.issue_d = pd.to_datetime(dfdates.issue_d)\n",
    "dfdates.earliest_cr_line = pd.to_datetime(dfdates.earliest_cr_line)\n",
    "\n",
    "# Today's datetime\n",
    "today = datetime.date.today()\n",
    "\n",
    "# Turn datetimes into elapsed months since the dates to today using monthmod.py\n",
    "dfdates.issue_d = dfdates.issue_d.apply(lambda x: (\n",
    "    monthmod(x, pd.Timestamp(today))[0].months))\n",
    "dfdates.earliest_cr_line = dfdates.earliest_cr_line.apply(lambda x: (\n",
    "    monthmod(x, pd.Timestamp(today))[0].months))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "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></th>\n",
       "      <th>issue_d</th>\n",
       "      <th>earliest_cr_line</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>2</th>\n",
       "      <td>28</td>\n",
       "      <td>145</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>28</td>\n",
       "      <td>175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>28</td>\n",
       "      <td>158</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>28</td>\n",
       "      <td>345</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>28</td>\n",
       "      <td>416</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       issue_d  earliest_cr_line\n",
       "Q1 2        28               145\n",
       "   4        28               175\n",
       "   8        28               158\n",
       "   13       28               345\n",
       "   17       28               416"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[dates] = dfdates.astype(int)\n",
    "df[dates].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Categorical Variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Categorical variables from previous analysis\n",
    "# \n",
    "cat = ['grade', 'sub_grade', 'home_ownership', 'verification_status', 'purpose']\n",
    "\n",
    "# One-hot encode categoricals\n",
    "dummified_cols = pd.get_dummies(df[cat])\n",
    "\n",
    "#Remove original columns and add encoded\n",
    "df.drop(cat, axis =1, inplace= True)\n",
    "df = df.join(dummified_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "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></th>\n",
       "      <th>grade_A</th>\n",
       "      <th>grade_B</th>\n",
       "      <th>grade_C</th>\n",
       "      <th>grade_D</th>\n",
       "      <th>grade_E</th>\n",
       "      <th>grade_F</th>\n",
       "      <th>grade_G</th>\n",
       "      <th>sub_grade_A1</th>\n",
       "      <th>sub_grade_A2</th>\n",
       "      <th>sub_grade_A3</th>\n",
       "      <th>...</th>\n",
       "      <th>purpose_home_improvement</th>\n",
       "      <th>purpose_house</th>\n",
       "      <th>purpose_major_purchase</th>\n",
       "      <th>purpose_medical</th>\n",
       "      <th>purpose_moving</th>\n",
       "      <th>purpose_other</th>\n",
       "      <th>purpose_renewable_energy</th>\n",
       "      <th>purpose_small_business</th>\n",
       "      <th>purpose_vacation</th>\n",
       "      <th>purpose_wedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">Q1</th>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 62 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       grade_A  grade_B  grade_C  grade_D  grade_E  grade_F  grade_G  \\\n",
       "Q1 2         0        1        0        0        0        0        0   \n",
       "   4         0        1        0        0        0        0        0   \n",
       "   8         0        0        1        0        0        0        0   \n",
       "   13        0        1        0        0        0        0        0   \n",
       "   17        0        1        0        0        0        0        0   \n",
       "\n",
       "       sub_grade_A1  sub_grade_A2  sub_grade_A3       ...         \\\n",
       "Q1 2              0             0             0       ...          \n",
       "   4              0             0             0       ...          \n",
       "   8              0             0             0       ...          \n",
       "   13             0             0             0       ...          \n",
       "   17             0             0             0       ...          \n",
       "\n",
       "       purpose_home_improvement  purpose_house  purpose_major_purchase  \\\n",
       "Q1 2                          0              0                       0   \n",
       "   4                          1              0                       0   \n",
       "   8                          1              0                       0   \n",
       "   13                         0              0                       0   \n",
       "   17                         0              0                       0   \n",
       "\n",
       "       purpose_medical  purpose_moving  purpose_other  \\\n",
       "Q1 2                 0               0              0   \n",
       "   4                 0               0              0   \n",
       "   8                 0               0              0   \n",
       "   13                0               0              0   \n",
       "   17                0               0              1   \n",
       "\n",
       "       purpose_renewable_energy  purpose_small_business  purpose_vacation  \\\n",
       "Q1 2                          0                       0                 0   \n",
       "   4                          0                       0                 0   \n",
       "   8                          0                       0                 0   \n",
       "   13                         0                       0                 0   \n",
       "   17                         0                       0                 0   \n",
       "\n",
       "       purpose_wedding  \n",
       "Q1 2                 0  \n",
       "   4                 0  \n",
       "   8                 0  \n",
       "   13                0  \n",
       "   17                0  \n",
       "\n",
       "[5 rows x 62 columns]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dummified_cols.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Missing Data\n",
    "Even after removing columns with more that 50% NaN, we're left with some missing data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    " # NaN entries? remove? equal zero? mean?  -> remove if NaN > 60, else median"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mths_since_last_delinq    92921\n",
      "il_util                   24401\n",
      "mths_since_recent_inq     16472\n",
      "num_tl_120dpd_2m           9356\n",
      "mths_since_rcnt_il         4567\n",
      "mo_sin_old_il_acct         4541\n",
      "bc_util                    2398\n",
      "percent_bc_gt_75           2299\n",
      "bc_open_to_buy             2285\n",
      "mths_since_recent_bc       2132\n",
      "revol_util                  129\n",
      "all_util                     44\n",
      "total_cu_tl                  30\n",
      "open_acc_6m                  30\n",
      "inq_last_12m                 30\n",
      "max_bal_bc                   29\n",
      "open_rv_24m                  29\n",
      "open_rv_12m                  29\n",
      "total_bal_il                 29\n",
      "inq_fi                       29\n",
      "dtype: int64\n",
      "Only showing first 20\n"
     ]
    }
   ],
   "source": [
    "print(df.isnull().sum().sort_values(ascending=False)[:20])\n",
    "print('Only showing first 20')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By removing the top ten columns that still have a lot of missing data, we can then input the missing data remaining without adding too much bias, or remove it without losing too many samples. This can be revisited if the results are not satisfactory enough and we think this columns could help to improve them. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['mths_since_last_delinq','il_util', 'mths_since_recent_inq', 'num_tl_120dpd_2m', 'mths_since_rcnt_il',\n",
    "         'mo_sin_old_il_acct', 'bc_util', 'percent_bc_gt_75', 'bc_open_to_buy', 'mths_since_recent_bc'], axis = 1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's remove the rows with the remaining NaN entries. We could also fill them with the mean, but removing them we don't introduce any bias."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# df.fillna(df.mean())\n",
    "df.dropna(axis=0, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Target Variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Removed column(s) = 92 - 127 = -35\n"
     ]
    }
   ],
   "source": [
    "# Remove target variable\n",
    "\n",
    "y = df['loan_status']\n",
    "df.drop(['loan_status'], axis = 1, inplace = True)\n",
    "print('Removed column(s) = 92 - {} = {}'.format(df.shape[1], (92-df.shape[1])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# turn target variable to binary \n",
    "y = pd.Series(y).str.replace('Charged Off', '1')\n",
    "y = pd.Series(y).str.replace('Fully Paid', '0').astype(int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train and Test Sets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can now split the data set in training and testing sets. The testing set will be the data referent to the Q4 index. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "names = df.columns.values\n",
    "X_test = df.loc['Q4', :]\n",
    "X_train = df.drop('Q4')\n",
    "y_test = y.loc['Q4', :]\n",
    "y_train = y.drop('Q4')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Standardize variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "def standardize (X):\n",
    "    scaler = StandardScaler()\n",
    "    X_norm = scaler.fit_transform(X)\n",
    "    X_norm = pd.DataFrame(data=X_norm, columns=names)\n",
    "    return X_norm, scaler\n",
    "\n",
    "X_train, Xzscaler = standardize(X_train)\n",
    "X_test = Xzscaler.transform(X_test) # the scaler modeled to the train set is applied on the test set\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Selection\n",
    "\n",
    "I'll experiment with a few different classification models. Logistic regression, Random Forest and GradientBoost which are know for having good performances. Also, by experimenting with different techniques (regression, trees, boosting) it is possible to conclude what kind better fits the problem. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to run the models and evaluate the results\n",
    "def run_models(X_train, y_train, X_test, y_test):\n",
    "    \n",
    "    clfs = {'LogisticRegression' : LogisticRegression(),\n",
    "            #'DecisionTree': DecisionTreeClassifier(random_state=1),\n",
    "            #'KNNClassifier': KNeighborsClassifier(n_neighbors=10),\n",
    "            'GradientBoosting': GradientBoostingClassifier(max_depth= 6, n_estimators=100, max_features = 0.3, random_state=1),\n",
    "            'RandomForestClassifier': RandomForestClassifier(n_estimators=10, random_state=1)\n",
    "            }\n",
    "    cols = ['model','matthews_corrcoef', 'roc_auc_score', 'precision_score', 'recall_score','f1_score']\n",
    "\n",
    "    models_report = pd.DataFrame(columns = cols)\n",
    "    conf_matrix = dict()\n",
    "\n",
    "    for clf, clf_name in zip(clfs.values(), clfs.keys()):\n",
    "\n",
    "        clf.fit(X_train, y_train)\n",
    "\n",
    "        y_pred = clf.predict(X_test)\n",
    "        y_score = clf.predict_proba(X_test)[:,1]\n",
    "\n",
    "        print('computing {}'.format(clf_name))\n",
    "\n",
    "        tmp = pd.Series({'model': clf_name,\n",
    "                         'roc_auc_score' : metrics.roc_auc_score(y_test, y_score),\n",
    "                         'matthews_corrcoef': metrics.matthews_corrcoef(y_test, y_pred),\n",
    "                         'precision_score': metrics.precision_score(y_test, y_pred),\n",
    "                         'recall_score': metrics.recall_score(y_test, y_pred),\n",
    "                         'f1_score': metrics.f1_score(y_test, y_pred)})\n",
    "\n",
    "        models_report = models_report.append(tmp, ignore_index = True)\n",
    "        conf_matrix[clf_name] = pd.crosstab(y_test, y_pred, rownames=['True'], colnames= ['Predicted'], margins=False)\n",
    "        fpr, tpr, thresholds = metrics.roc_curve(y_test, y_score, drop_intermediate = False, pos_label = 1)\n",
    "\n",
    "        plt.figure(1, figsize=(6,6))\n",
    "        plt.xlabel('false positive rate')\n",
    "        plt.ylabel('true positive rate')\n",
    "        plt.title('ROC curve')\n",
    "        plt.plot(fpr, tpr, label = clf_name )\n",
    "        plt.legend(loc=4, prop={'size':11})\n",
    "    plt.plot([0,1],[0,1], color = 'black')\n",
    "    \n",
    "    return models_report, conf_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "computing LogisticRegression\n",
      "computing GradientBoosting\n",
      "computing RandomForestClassifier\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# this step takes a while to process.\n",
    "# Be patient the models are being generated, predictions made and results evaluated. \n",
    "\n",
    "models_report, conf_matrix = run_models(X_train, y_train, X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "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>model</th>\n",
       "      <th>matthews_corrcoef</th>\n",
       "      <th>roc_auc_score</th>\n",
       "      <th>precision_score</th>\n",
       "      <th>recall_score</th>\n",
       "      <th>f1_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>LogisticRegression</td>\n",
       "      <td>0.995379</td>\n",
       "      <td>0.999383</td>\n",
       "      <td>0.999884</td>\n",
       "      <td>0.992959</td>\n",
       "      <td>0.996409</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GradientBoosting</td>\n",
       "      <td>0.999180</td>\n",
       "      <td>0.999927</td>\n",
       "      <td>0.999769</td>\n",
       "      <td>0.998961</td>\n",
       "      <td>0.999365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RandomForestClassifier</td>\n",
       "      <td>0.985064</td>\n",
       "      <td>0.999653</td>\n",
       "      <td>0.993354</td>\n",
       "      <td>0.983493</td>\n",
       "      <td>0.988399</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    model  matthews_corrcoef  roc_auc_score  precision_score  \\\n",
       "0      LogisticRegression           0.995379       0.999383         0.999884   \n",
       "1        GradientBoosting           0.999180       0.999927         0.999769   \n",
       "2  RandomForestClassifier           0.985064       0.999653         0.993354   \n",
       "\n",
       "   recall_score  f1_score  \n",
       "0      0.992959  0.996409  \n",
       "1      0.998961  0.999365  \n",
       "2      0.983493  0.988399  "
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models_report"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Final Notes: \n",
    "    - We could also do another split for cross validation and hyperparameter tunning, which could improve the performance of the models however this is time consuming and I considered it to be out of the scope of this challenge. Besides: the performances are outstanding already. "
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}