predict_and_rank.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict and rank NPLs\n",
    "In this notebook, I apply the predictive models developed in the *modeling_and_performance_part2.ipynb* over the most recent NPLs (09-2017) in order to predict the **probability of recovery** and the **expected recovery rate** in the following 12 months. The NPLs are then ranked by those two predicted values.\n",
    "\n",
    "In order to facilitate the work of portfolio managers, the probability of recovery is used to define a **recovery class**, i.e. group of NPLs within a given range of probability. The following class are defined:\n",
    "\n",
    "* Class A: very likely to recover\n",
    "    * Probability of recovery [0.75, 1.00];\n",
    "* Class B: likely to recover\n",
    "    * Probability of recovery [0.50, 0.75];\n",
    "* Class C: unlikely to recover:\n",
    "    * Probability of recovery [0.25, 0.50];\n",
    "* Class D: very unlikely to recover\n",
    "    * Probability of recovery [0.00, 0.25];\n",
    "\n",
    "The final results will then be summarised in term of recovery class, providing the total recovery rate and the total recovered amount expected in the 12 months following the 09-2017.\n",
    "\n",
    "The ranked list of NPL will be make aviable on a csv file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import os.path as path\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.dates as mdates\n",
    "from classifier_utils import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load pre-processed dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('sample_for_model.csv'\n",
    "                  , parse_dates=['NUM_ANNO', 'NUM_YYYYMM', 'NUM_DATA_SOFFERENZA']\n",
    "                  , dtype={'HAS_RECOVERED_12M': bool, 'HAS_RECOVERED_12M_AHEAD': bool})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(484743, 46)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "numerical_cols=[\n",
    "    'IMP_GBV',\n",
    "    'IMP_CUSUM_INCASSI',\n",
    "    'MAX_IMP_GBV',\n",
    "    'RECOVERY_RATE_TOTAL',\n",
    "    'NUM_MONTHS_SOFFERENZA',\n",
    "    'Contabile linea interessi di mora',\n",
    "    'Credito vantato',\n",
    "    'NUM_ENTI_SEGNALANTI',\n",
    "    'NUM_COMPONENTI_COINTESTAZIONE',\n",
    "    'NUM_ID_RAPPORTO',\n",
    "    'NUM_ID_GARANTE',\n",
    "    'IMP_INCASSI_ROLLING_12M'\n",
    "    ,'IMP_GAR',\n",
    "]\n",
    "categorical_cols=[]\n",
    "\n",
    "target_classifier = 'HAS_RECOVERED_12M_AHEAD'\n",
    "target_regressor = 'RECOVERY_RATE_12M_AHEAD'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Encode categorical features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This is not necessary with this set of feature, but \n",
    "# it mightr be required in future release.\n",
    "prefix = ['CAT_' + s for s in categorical_cols]\n",
    "\n",
    "df_encoded=pd.get_dummies(df, columns=categorical_cols, dummy_na=True, prefix=prefix)\n",
    "\n",
    "encoded_categorical_cols = list(filter(lambda x: 'CAT_' in x, df_encoded.columns))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(484743, 46)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_encoded.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prediction sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(18786, 46)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_prediction=df_encoded.loc[df_encoded['NUM_YYYYMM'] == df_encoded['NUM_YYYYMM'].max()]\n",
    "sample_prediction.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load models\n",
    "Load model from the external joblib files and prepare the function to apply them to the different datapoints depending from the customer type."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "\n",
    "models_dict = {}\n",
    "\n",
    "cols = numerical_cols + encoded_categorical_cols\n",
    "\n",
    "models_dict['F'] = {\n",
    "    'features' : cols,\n",
    "    'classifier' : joblib.load('classifier_random_forest_pf_20191105.joblib'),\n",
    "    'regressor' : joblib.load('regressor_random_forest_pf_20191105.joblib')\n",
    "}\n",
    "\n",
    "models_dict['G'] = {\n",
    "    'features' : list(filter(lambda s: s!= 'IMP_GAR', cols)),\n",
    "    'classifier' : joblib.load('classifier_random_forest_pg_20191105.joblib'),\n",
    "    'regressor' : joblib.load('regressor_random_forest_pg_20191105.joblib')\n",
    "}\n",
    "\n",
    "models_dict['C'] = {\n",
    "    'features' : list(filter(lambda s: s!= 'IMP_GAR', cols)),\n",
    "    'classifier' : joblib.load('classifier_random_forest_c_20191105.joblib'),\n",
    "    'regressor' : joblib.load('regressor_random_forest_c_20191105.joblib')\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def apply_model(name, df, models_dict=models_dict):\n",
    "    \n",
    "    prob = models_dict[name]['classifier'].predict_proba(df[models_dict[name]['features']])\n",
    "    pred = models_dict[name]['regressor'].predict(df[models_dict[name]['features']])\n",
    "\n",
    "    return pd.DataFrame({\n",
    "        'ID_CUSTOMER' : df['ID_CUSTOMER'].values,\n",
    "        'IMP_GBV' : df['IMP_GBV'].values,\n",
    "        'PRED_PROB_RECOVERY_12M_AHEAD' : prob[:,1], # I am only interested in the probability for class 1, i.e. target variable == True    \n",
    "        'PRED_RECOVERY_RATE_12M_AHEAD' : pred\n",
    "    })"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Apply model "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_prediction=sample_prediction.groupby('COD_TIPO_NDG').apply(\n",
    "    lambda x: apply_model(x.name, x)\n",
    ").reset_index(-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>COD_TIPO_NDG</th>\n",
       "      <th>ID_CUSTOMER</th>\n",
       "      <th>IMP_GBV</th>\n",
       "      <th>PRED_PROB_RECOVERY_12M_AHEAD</th>\n",
       "      <th>PRED_RECOVERY_RATE_12M_AHEAD</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>C</td>\n",
       "      <td>7</td>\n",
       "      <td>63.13</td>\n",
       "      <td>0.577436</td>\n",
       "      <td>0.504080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>C</td>\n",
       "      <td>11</td>\n",
       "      <td>1930.97</td>\n",
       "      <td>0.856599</td>\n",
       "      <td>0.415380</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>C</td>\n",
       "      <td>59</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.058328</td>\n",
       "      <td>0.003402</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>C</td>\n",
       "      <td>78</td>\n",
       "      <td>759.09</td>\n",
       "      <td>0.037779</td>\n",
       "      <td>0.032719</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>C</td>\n",
       "      <td>91</td>\n",
       "      <td>105.81</td>\n",
       "      <td>0.228014</td>\n",
       "      <td>0.159764</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  COD_TIPO_NDG  ID_CUSTOMER  IMP_GBV  PRED_PROB_RECOVERY_12M_AHEAD  \\\n",
       "0            C            7    63.13                      0.577436   \n",
       "1            C           11  1930.97                      0.856599   \n",
       "2            C           59    76.10                      0.058328   \n",
       "3            C           78   759.09                      0.037779   \n",
       "4            C           91   105.81                      0.228014   \n",
       "\n",
       "   PRED_RECOVERY_RATE_12M_AHEAD  \n",
       "0                      0.504080  \n",
       "1                      0.415380  \n",
       "2                      0.003402  \n",
       "3                      0.032719  \n",
       "4                      0.159764  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_prediction.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute the predicted recovery amount\n",
    "model_prediction['PRED_INCASSI_12M_AHEAD'] = model_prediction['PRED_RECOVERY_RATE_12M_AHEAD'] * model_prediction['IMP_GBV']\n",
    "\n",
    "# Bucket the probability into 4 different class\n",
    "model_prediction['PRED_RECOVERY_CLASS'] = pd.cut(\n",
    "    model_prediction['PRED_PROB_RECOVERY_12M_AHEAD'], \n",
    "    bins=[0, 0.25,0.5,0.75,1.], \n",
    "    labels=['D', 'C', 'B', 'A'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>COD_TIPO_NDG</th>\n",
       "      <th>ID_CUSTOMER</th>\n",
       "      <th>IMP_GBV</th>\n",
       "      <th>PRED_PROB_RECOVERY_12M_AHEAD</th>\n",
       "      <th>PRED_RECOVERY_RATE_12M_AHEAD</th>\n",
       "      <th>PRED_INCASSI_12M_AHEAD</th>\n",
       "      <th>PRED_RECOVERY_CLASS</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>C</td>\n",
       "      <td>7</td>\n",
       "      <td>63.13</td>\n",
       "      <td>0.577436</td>\n",
       "      <td>0.504080</td>\n",
       "      <td>31.822598</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>C</td>\n",
       "      <td>11</td>\n",
       "      <td>1930.97</td>\n",
       "      <td>0.856599</td>\n",
       "      <td>0.415380</td>\n",
       "      <td>802.085882</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>C</td>\n",
       "      <td>59</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.058328</td>\n",
       "      <td>0.003402</td>\n",
       "      <td>0.258873</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>C</td>\n",
       "      <td>78</td>\n",
       "      <td>759.09</td>\n",
       "      <td>0.037779</td>\n",
       "      <td>0.032719</td>\n",
       "      <td>24.836991</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>C</td>\n",
       "      <td>91</td>\n",
       "      <td>105.81</td>\n",
       "      <td>0.228014</td>\n",
       "      <td>0.159764</td>\n",
       "      <td>16.904667</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  COD_TIPO_NDG  ID_CUSTOMER  IMP_GBV  PRED_PROB_RECOVERY_12M_AHEAD  \\\n",
       "0            C            7    63.13                      0.577436   \n",
       "1            C           11  1930.97                      0.856599   \n",
       "2            C           59    76.10                      0.058328   \n",
       "3            C           78   759.09                      0.037779   \n",
       "4            C           91   105.81                      0.228014   \n",
       "\n",
       "   PRED_RECOVERY_RATE_12M_AHEAD  PRED_INCASSI_12M_AHEAD PRED_RECOVERY_CLASS  \n",
       "0                      0.504080               31.822598                   B  \n",
       "1                      0.415380              802.085882                   A  \n",
       "2                      0.003402                0.258873                   D  \n",
       "3                      0.032719               24.836991                   D  \n",
       "4                      0.159764               16.904667                   D  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_prediction.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sorting NPL by the probability of recovery and by the recovery rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>COD_TIPO_NDG</th>\n",
       "      <th>ID_CUSTOMER</th>\n",
       "      <th>IMP_GBV</th>\n",
       "      <th>PRED_PROB_RECOVERY_12M_AHEAD</th>\n",
       "      <th>PRED_RECOVERY_RATE_12M_AHEAD</th>\n",
       "      <th>PRED_INCASSI_12M_AHEAD</th>\n",
       "      <th>PRED_RECOVERY_CLASS</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4990</th>\n",
       "      <td>C</td>\n",
       "      <td>27024</td>\n",
       "      <td>517.45</td>\n",
       "      <td>0.830976</td>\n",
       "      <td>7.616390e-01</td>\n",
       "      <td>394.110118</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3075</th>\n",
       "      <td>F</td>\n",
       "      <td>17405</td>\n",
       "      <td>3032.61</td>\n",
       "      <td>0.896712</td>\n",
       "      <td>6.793517e-01</td>\n",
       "      <td>2060.208666</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138</th>\n",
       "      <td>G</td>\n",
       "      <td>434</td>\n",
       "      <td>283.18</td>\n",
       "      <td>0.883208</td>\n",
       "      <td>6.741502e-01</td>\n",
       "      <td>190.905846</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>385</th>\n",
       "      <td>C</td>\n",
       "      <td>6735</td>\n",
       "      <td>467.58</td>\n",
       "      <td>0.948984</td>\n",
       "      <td>6.318838e-01</td>\n",
       "      <td>295.456212</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>373</th>\n",
       "      <td>C</td>\n",
       "      <td>6706</td>\n",
       "      <td>500.14</td>\n",
       "      <td>0.935252</td>\n",
       "      <td>6.190703e-01</td>\n",
       "      <td>309.621822</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5753</th>\n",
       "      <td>F</td>\n",
       "      <td>27177</td>\n",
       "      <td>8210.79</td>\n",
       "      <td>0.888776</td>\n",
       "      <td>6.135325e-01</td>\n",
       "      <td>5037.586623</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>630</th>\n",
       "      <td>F</td>\n",
       "      <td>6118</td>\n",
       "      <td>812.93</td>\n",
       "      <td>0.827547</td>\n",
       "      <td>5.980373e-01</td>\n",
       "      <td>486.162472</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1453</th>\n",
       "      <td>G</td>\n",
       "      <td>7161</td>\n",
       "      <td>1596.68</td>\n",
       "      <td>0.797899</td>\n",
       "      <td>5.955661e-01</td>\n",
       "      <td>950.928505</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3314</th>\n",
       "      <td>F</td>\n",
       "      <td>18357</td>\n",
       "      <td>771.51</td>\n",
       "      <td>0.904660</td>\n",
       "      <td>5.904105e-01</td>\n",
       "      <td>455.507616</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5645</th>\n",
       "      <td>F</td>\n",
       "      <td>26852</td>\n",
       "      <td>1395.38</td>\n",
       "      <td>0.788217</td>\n",
       "      <td>5.773679e-01</td>\n",
       "      <td>805.647588</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3971</th>\n",
       "      <td>F</td>\n",
       "      <td>21253</td>\n",
       "      <td>848.25</td>\n",
       "      <td>0.910101</td>\n",
       "      <td>5.738557e-01</td>\n",
       "      <td>486.773088</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2643</th>\n",
       "      <td>F</td>\n",
       "      <td>15530</td>\n",
       "      <td>842.37</td>\n",
       "      <td>0.947028</td>\n",
       "      <td>5.716828e-01</td>\n",
       "      <td>481.568401</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1704</th>\n",
       "      <td>C</td>\n",
       "      <td>14608</td>\n",
       "      <td>1013.24</td>\n",
       "      <td>0.942002</td>\n",
       "      <td>5.684745e-01</td>\n",
       "      <td>576.001063</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>C</td>\n",
       "      <td>3074</td>\n",
       "      <td>708.52</td>\n",
       "      <td>0.890260</td>\n",
       "      <td>5.605339e-01</td>\n",
       "      <td>397.149451</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5330</th>\n",
       "      <td>F</td>\n",
       "      <td>25850</td>\n",
       "      <td>640.92</td>\n",
       "      <td>0.908300</td>\n",
       "      <td>5.577959e-01</td>\n",
       "      <td>357.502548</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5589</th>\n",
       "      <td>G</td>\n",
       "      <td>21781</td>\n",
       "      <td>1328.59</td>\n",
       "      <td>0.872239</td>\n",
       "      <td>5.547559e-01</td>\n",
       "      <td>737.043190</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1782</th>\n",
       "      <td>F</td>\n",
       "      <td>11930</td>\n",
       "      <td>1027.68</td>\n",
       "      <td>0.927092</td>\n",
       "      <td>5.460356e-01</td>\n",
       "      <td>561.149842</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4869</th>\n",
       "      <td>F</td>\n",
       "      <td>24438</td>\n",
       "      <td>819.21</td>\n",
       "      <td>0.920173</td>\n",
       "      <td>5.364831e-01</td>\n",
       "      <td>439.492309</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5183</th>\n",
       "      <td>C</td>\n",
       "      <td>27607</td>\n",
       "      <td>498.11</td>\n",
       "      <td>0.855807</td>\n",
       "      <td>5.358900e-01</td>\n",
       "      <td>266.932177</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2127</th>\n",
       "      <td>C</td>\n",
       "      <td>16600</td>\n",
       "      <td>909.70</td>\n",
       "      <td>0.961952</td>\n",
       "      <td>5.355978e-01</td>\n",
       "      <td>487.233343</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2506</th>\n",
       "      <td>F</td>\n",
       "      <td>15016</td>\n",
       "      <td>1123.77</td>\n",
       "      <td>0.781803</td>\n",
       "      <td>5.291172e-01</td>\n",
       "      <td>594.606023</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5236</th>\n",
       "      <td>F</td>\n",
       "      <td>25572</td>\n",
       "      <td>1164.71</td>\n",
       "      <td>0.799099</td>\n",
       "      <td>5.289858e-01</td>\n",
       "      <td>616.115040</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1971</th>\n",
       "      <td>C</td>\n",
       "      <td>15662</td>\n",
       "      <td>539.63</td>\n",
       "      <td>0.910532</td>\n",
       "      <td>5.265349e-01</td>\n",
       "      <td>284.134019</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1787</th>\n",
       "      <td>G</td>\n",
       "      <td>8290</td>\n",
       "      <td>335.89</td>\n",
       "      <td>0.787508</td>\n",
       "      <td>5.230102e-01</td>\n",
       "      <td>175.673887</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1608</th>\n",
       "      <td>F</td>\n",
       "      <td>11257</td>\n",
       "      <td>1221.49</td>\n",
       "      <td>0.837005</td>\n",
       "      <td>5.120363e-01</td>\n",
       "      <td>625.447267</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>C</td>\n",
       "      <td>586</td>\n",
       "      <td>954.90</td>\n",
       "      <td>0.753847</td>\n",
       "      <td>5.119777e-01</td>\n",
       "      <td>488.887534</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4071</th>\n",
       "      <td>G</td>\n",
       "      <td>16408</td>\n",
       "      <td>1438.53</td>\n",
       "      <td>0.961050</td>\n",
       "      <td>5.102352e-01</td>\n",
       "      <td>733.988627</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5858</th>\n",
       "      <td>G</td>\n",
       "      <td>22731</td>\n",
       "      <td>1750.39</td>\n",
       "      <td>0.857310</td>\n",
       "      <td>5.093492e-01</td>\n",
       "      <td>891.559737</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1682</th>\n",
       "      <td>G</td>\n",
       "      <td>7943</td>\n",
       "      <td>4133.09</td>\n",
       "      <td>0.787591</td>\n",
       "      <td>5.078813e-01</td>\n",
       "      <td>2099.119323</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3903</th>\n",
       "      <td>F</td>\n",
       "      <td>21009</td>\n",
       "      <td>909.84</td>\n",
       "      <td>0.920168</td>\n",
       "      <td>5.057613e-01</td>\n",
       "      <td>460.161887</td>\n",
       "      <td>A</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>900</th>\n",
       "      <td>G</td>\n",
       "      <td>4371</td>\n",
       "      <td>994.73</td>\n",
       "      <td>0.025616</td>\n",
       "      <td>4.689647e-08</td>\n",
       "      <td>0.000047</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1245</th>\n",
       "      <td>C</td>\n",
       "      <td>12384</td>\n",
       "      <td>13371.86</td>\n",
       "      <td>0.007695</td>\n",
       "      <td>3.991368e-08</td>\n",
       "      <td>0.000534</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1455</th>\n",
       "      <td>F</td>\n",
       "      <td>10586</td>\n",
       "      <td>8983.91</td>\n",
       "      <td>0.019627</td>\n",
       "      <td>3.968644e-08</td>\n",
       "      <td>0.000357</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4916</th>\n",
       "      <td>G</td>\n",
       "      <td>19381</td>\n",
       "      <td>334.49</td>\n",
       "      <td>0.003509</td>\n",
       "      <td>7.642091e-09</td>\n",
       "      <td>0.000003</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>C</td>\n",
       "      <td>1325</td>\n",
       "      <td>123.17</td>\n",
       "      <td>0.013524</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>394</th>\n",
       "      <td>C</td>\n",
       "      <td>6750</td>\n",
       "      <td>4272.77</td>\n",
       "      <td>0.025791</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>689</th>\n",
       "      <td>C</td>\n",
       "      <td>8878</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.008624</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>708</th>\n",
       "      <td>C</td>\n",
       "      <td>9132</td>\n",
       "      <td>14891.58</td>\n",
       "      <td>0.018548</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1272</th>\n",
       "      <td>C</td>\n",
       "      <td>12519</td>\n",
       "      <td>11393.73</td>\n",
       "      <td>0.018282</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1313</th>\n",
       "      <td>C</td>\n",
       "      <td>12702</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.007103</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1314</th>\n",
       "      <td>C</td>\n",
       "      <td>12713</td>\n",
       "      <td>78.04</td>\n",
       "      <td>0.017725</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1511</th>\n",
       "      <td>C</td>\n",
       "      <td>13771</td>\n",
       "      <td>1586.39</td>\n",
       "      <td>0.055894</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1672</th>\n",
       "      <td>C</td>\n",
       "      <td>14505</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.003280</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1675</th>\n",
       "      <td>C</td>\n",
       "      <td>14515</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.004898</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1679</th>\n",
       "      <td>C</td>\n",
       "      <td>14526</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.002109</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1681</th>\n",
       "      <td>C</td>\n",
       "      <td>14530</td>\n",
       "      <td>76.10</td>\n",
       "      <td>0.003893</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1833</th>\n",
       "      <td>C</td>\n",
       "      <td>15293</td>\n",
       "      <td>3986.29</td>\n",
       "      <td>0.008354</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3242</th>\n",
       "      <td>C</td>\n",
       "      <td>21628</td>\n",
       "      <td>801.76</td>\n",
       "      <td>0.018245</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4451</th>\n",
       "      <td>C</td>\n",
       "      <td>25483</td>\n",
       "      <td>247.92</td>\n",
       "      <td>0.019978</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4860</th>\n",
       "      <td>C</td>\n",
       "      <td>26584</td>\n",
       "      <td>182.25</td>\n",
       "      <td>0.072508</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5271</th>\n",
       "      <td>C</td>\n",
       "      <td>27872</td>\n",
       "      <td>126.14</td>\n",
       "      <td>0.013751</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54</th>\n",
       "      <td>F</td>\n",
       "      <td>304</td>\n",
       "      <td>989.62</td>\n",
       "      <td>0.016373</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71</th>\n",
       "      <td>F</td>\n",
       "      <td>363</td>\n",
       "      <td>1169.35</td>\n",
       "      <td>0.013024</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>204</th>\n",
       "      <td>F</td>\n",
       "      <td>2500</td>\n",
       "      <td>984.32</td>\n",
       "      <td>0.012481</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>113</th>\n",
       "      <td>G</td>\n",
       "      <td>378</td>\n",
       "      <td>792.40</td>\n",
       "      <td>0.032849</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>337</th>\n",
       "      <td>G</td>\n",
       "      <td>1086</td>\n",
       "      <td>1566.67</td>\n",
       "      <td>0.016771</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>413</th>\n",
       "      <td>G</td>\n",
       "      <td>1406</td>\n",
       "      <td>623.84</td>\n",
       "      <td>0.018448</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2245</th>\n",
       "      <td>G</td>\n",
       "      <td>9946</td>\n",
       "      <td>425.03</td>\n",
       "      <td>0.006148</td>\n",
       "      <td>0.000000e+00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>D</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4017</th>\n",
       "      <td>G</td>\n",
       "      <td>16204</td>\n",
       "      <td>9167.09</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3.846357e-03</td>\n",
       "      <td>35.259904</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1333</th>\n",
       "      <td>G</td>\n",
       "      <td>6677</td>\n",
       "      <td>15575.00</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.585260e-06</td>\n",
       "      <td>0.040265</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>18786 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     COD_TIPO_NDG  ID_CUSTOMER   IMP_GBV  PRED_PROB_RECOVERY_12M_AHEAD  \\\n",
       "4990            C        27024    517.45                      0.830976   \n",
       "3075            F        17405   3032.61                      0.896712   \n",
       "138             G          434    283.18                      0.883208   \n",
       "385             C         6735    467.58                      0.948984   \n",
       "373             C         6706    500.14                      0.935252   \n",
       "5753            F        27177   8210.79                      0.888776   \n",
       "630             F         6118    812.93                      0.827547   \n",
       "1453            G         7161   1596.68                      0.797899   \n",
       "3314            F        18357    771.51                      0.904660   \n",
       "5645            F        26852   1395.38                      0.788217   \n",
       "3971            F        21253    848.25                      0.910101   \n",
       "2643            F        15530    842.37                      0.947028   \n",
       "1704            C        14608   1013.24                      0.942002   \n",
       "60              C         3074    708.52                      0.890260   \n",
       "5330            F        25850    640.92                      0.908300   \n",
       "5589            G        21781   1328.59                      0.872239   \n",
       "1782            F        11930   1027.68                      0.927092   \n",
       "4869            F        24438    819.21                      0.920173   \n",
       "5183            C        27607    498.11                      0.855807   \n",
       "2127            C        16600    909.70                      0.961952   \n",
       "2506            F        15016   1123.77                      0.781803   \n",
       "5236            F        25572   1164.71                      0.799099   \n",
       "1971            C        15662    539.63                      0.910532   \n",
       "1787            G         8290    335.89                      0.787508   \n",
       "1608            F        11257   1221.49                      0.837005   \n",
       "18              C          586    954.90                      0.753847   \n",
       "4071            G        16408   1438.53                      0.961050   \n",
       "5858            G        22731   1750.39                      0.857310   \n",
       "1682            G         7943   4133.09                      0.787591   \n",
       "3903            F        21009    909.84                      0.920168   \n",
       "...           ...          ...       ...                           ...   \n",
       "900             G         4371    994.73                      0.025616   \n",
       "1245            C        12384  13371.86                      0.007695   \n",
       "1455            F        10586   8983.91                      0.019627   \n",
       "4916            G        19381    334.49                      0.003509   \n",
       "39              C         1325    123.17                      0.013524   \n",
       "394             C         6750   4272.77                      0.025791   \n",
       "689             C         8878     76.10                      0.008624   \n",
       "708             C         9132  14891.58                      0.018548   \n",
       "1272            C        12519  11393.73                      0.018282   \n",
       "1313            C        12702     76.10                      0.007103   \n",
       "1314            C        12713     78.04                      0.017725   \n",
       "1511            C        13771   1586.39                      0.055894   \n",
       "1672            C        14505     76.10                      0.003280   \n",
       "1675            C        14515     76.10                      0.004898   \n",
       "1679            C        14526     76.10                      0.002109   \n",
       "1681            C        14530     76.10                      0.003893   \n",
       "1833            C        15293   3986.29                      0.008354   \n",
       "3242            C        21628    801.76                      0.018245   \n",
       "4451            C        25483    247.92                      0.019978   \n",
       "4860            C        26584    182.25                      0.072508   \n",
       "5271            C        27872    126.14                      0.013751   \n",
       "54              F          304    989.62                      0.016373   \n",
       "71              F          363   1169.35                      0.013024   \n",
       "204             F         2500    984.32                      0.012481   \n",
       "113             G          378    792.40                      0.032849   \n",
       "337             G         1086   1566.67                      0.016771   \n",
       "413             G         1406    623.84                      0.018448   \n",
       "2245            G         9946    425.03                      0.006148   \n",
       "4017            G        16204   9167.09                      0.000000   \n",
       "1333            G         6677  15575.00                      0.000000   \n",
       "\n",
       "      PRED_RECOVERY_RATE_12M_AHEAD  PRED_INCASSI_12M_AHEAD PRED_RECOVERY_CLASS  \n",
       "4990                  7.616390e-01              394.110118                   A  \n",
       "3075                  6.793517e-01             2060.208666                   A  \n",
       "138                   6.741502e-01              190.905846                   A  \n",
       "385                   6.318838e-01              295.456212                   A  \n",
       "373                   6.190703e-01              309.621822                   A  \n",
       "5753                  6.135325e-01             5037.586623                   A  \n",
       "630                   5.980373e-01              486.162472                   A  \n",
       "1453                  5.955661e-01              950.928505                   A  \n",
       "3314                  5.904105e-01              455.507616                   A  \n",
       "5645                  5.773679e-01              805.647588                   A  \n",
       "3971                  5.738557e-01              486.773088                   A  \n",
       "2643                  5.716828e-01              481.568401                   A  \n",
       "1704                  5.684745e-01              576.001063                   A  \n",
       "60                    5.605339e-01              397.149451                   A  \n",
       "5330                  5.577959e-01              357.502548                   A  \n",
       "5589                  5.547559e-01              737.043190                   A  \n",
       "1782                  5.460356e-01              561.149842                   A  \n",
       "4869                  5.364831e-01              439.492309                   A  \n",
       "5183                  5.358900e-01              266.932177                   A  \n",
       "2127                  5.355978e-01              487.233343                   A  \n",
       "2506                  5.291172e-01              594.606023                   A  \n",
       "5236                  5.289858e-01              616.115040                   A  \n",
       "1971                  5.265349e-01              284.134019                   A  \n",
       "1787                  5.230102e-01              175.673887                   A  \n",
       "1608                  5.120363e-01              625.447267                   A  \n",
       "18                    5.119777e-01              488.887534                   A  \n",
       "4071                  5.102352e-01              733.988627                   A  \n",
       "5858                  5.093492e-01              891.559737                   A  \n",
       "1682                  5.078813e-01             2099.119323                   A  \n",
       "3903                  5.057613e-01              460.161887                   A  \n",
       "...                            ...                     ...                 ...  \n",
       "900                   4.689647e-08                0.000047                   D  \n",
       "1245                  3.991368e-08                0.000534                   D  \n",
       "1455                  3.968644e-08                0.000357                   D  \n",
       "4916                  7.642091e-09                0.000003                   D  \n",
       "39                    0.000000e+00                0.000000                   D  \n",
       "394                   0.000000e+00                0.000000                   D  \n",
       "689                   0.000000e+00                0.000000                   D  \n",
       "708                   0.000000e+00                0.000000                   D  \n",
       "1272                  0.000000e+00                0.000000                   D  \n",
       "1313                  0.000000e+00                0.000000                   D  \n",
       "1314                  0.000000e+00                0.000000                   D  \n",
       "1511                  0.000000e+00                0.000000                   D  \n",
       "1672                  0.000000e+00                0.000000                   D  \n",
       "1675                  0.000000e+00                0.000000                   D  \n",
       "1679                  0.000000e+00                0.000000                   D  \n",
       "1681                  0.000000e+00                0.000000                   D  \n",
       "1833                  0.000000e+00                0.000000                   D  \n",
       "3242                  0.000000e+00                0.000000                   D  \n",
       "4451                  0.000000e+00                0.000000                   D  \n",
       "4860                  0.000000e+00                0.000000                   D  \n",
       "5271                  0.000000e+00                0.000000                   D  \n",
       "54                    0.000000e+00                0.000000                   D  \n",
       "71                    0.000000e+00                0.000000                   D  \n",
       "204                   0.000000e+00                0.000000                   D  \n",
       "113                   0.000000e+00                0.000000                   D  \n",
       "337                   0.000000e+00                0.000000                   D  \n",
       "413                   0.000000e+00                0.000000                   D  \n",
       "2245                  0.000000e+00                0.000000                   D  \n",
       "4017                  3.846357e-03               35.259904                 NaN  \n",
       "1333                  2.585260e-06                0.040265                 NaN  \n",
       "\n",
       "[18786 rows x 7 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "npl_sorted = model_prediction.sort_values(by=['PRED_RECOVERY_CLASS', 'PRED_RECOVERY_RATE_12M_AHEAD'], ascending=False)\n",
    "display(npl_sorted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "npl_sorted.to_csv('npl_sorted.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* NPLs are sorted by the probability of recovery and by the recovery rate expected in the next 12 months.\n",
    "* The value of the probability is used to define a predicted *recovery class*: A, B, C, D\n",
    "* The predicted values are exported to a csv files together additional informatio that might be usefull."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Summary "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "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>NUM_CUSTOMERS</th>\n",
       "      <th>FRAC_CUSTOMERS</th>\n",
       "      <th>TOTAL_IMP_GBV</th>\n",
       "      <th>PRED_TOTAL_INCASSI</th>\n",
       "      <th>PRED_TOTAL_RECOVERY_RATE</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PRED_RECOVERY_CLASS</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>D</th>\n",
       "      <td>12153</td>\n",
       "      <td>0.646987</td>\n",
       "      <td>71896323.30</td>\n",
       "      <td>2.501808e+06</td>\n",
       "      <td>0.034797</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>C</th>\n",
       "      <td>3847</td>\n",
       "      <td>0.204802</td>\n",
       "      <td>25850577.60</td>\n",
       "      <td>1.953762e+06</td>\n",
       "      <td>0.075579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>B</th>\n",
       "      <td>1357</td>\n",
       "      <td>0.072242</td>\n",
       "      <td>8580449.95</td>\n",
       "      <td>8.459877e+05</td>\n",
       "      <td>0.098595</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A</th>\n",
       "      <td>1427</td>\n",
       "      <td>0.075969</td>\n",
       "      <td>7516635.67</td>\n",
       "      <td>7.698283e+05</td>\n",
       "      <td>0.102417</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     NUM_CUSTOMERS  FRAC_CUSTOMERS  TOTAL_IMP_GBV  \\\n",
       "PRED_RECOVERY_CLASS                                                 \n",
       "D                            12153        0.646987    71896323.30   \n",
       "C                             3847        0.204802    25850577.60   \n",
       "B                             1357        0.072242     8580449.95   \n",
       "A                             1427        0.075969     7516635.67   \n",
       "\n",
       "                     PRED_TOTAL_INCASSI  PRED_TOTAL_RECOVERY_RATE  \n",
       "PRED_RECOVERY_CLASS                                                \n",
       "D                          2.501808e+06                  0.034797  \n",
       "C                          1.953762e+06                  0.075579  \n",
       "B                          8.459877e+05                  0.098595  \n",
       "A                          7.698283e+05                  0.102417  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "summary = model_prediction.groupby('PRED_RECOVERY_CLASS').agg({\n",
    "    'PRED_RECOVERY_CLASS' : 'size',\n",
    "    'IMP_GBV' : 'sum',\n",
    "    'PRED_INCASSI_12M_AHEAD' : 'sum'\n",
    "})\n",
    "\n",
    "summary.rename(columns={\n",
    "    'PRED_RECOVERY_CLASS': 'NUM_CUSTOMERS',\n",
    "    'IMP_GBV' : 'TOTAL_IMP_GBV',\n",
    "    'PRED_INCASSI_12M_AHEAD' : 'PRED_TOTAL_INCASSI'    \n",
    "}, inplace=True)\n",
    "\n",
    "summary['FRAC_CUSTOMERS']= summary['NUM_CUSTOMERS']/summary['NUM_CUSTOMERS'].sum()\n",
    "\n",
    "summary['PRED_TOTAL_RECOVERY_RATE']=summary['PRED_TOTAL_INCASSI'] / summary['TOTAL_IMP_GBV']\n",
    "\n",
    "cols = ['NUM_CUSTOMERS', 'FRAC_CUSTOMERS', 'TOTAL_IMP_GBV', 'PRED_TOTAL_INCASSI', 'PRED_TOTAL_RECOVERY_RATE']\n",
    "\n",
    "display(summary[cols])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The table above summarise the level of recovery expected at 09-2017 from different recovery class. \n",
    "\n",
    "Class A contains the NPLs which are more likely to recover (Prob > 75%). A total of 1427 NPLs (7.6%) are in this class. The total debt corresponding to this class is 7.5M€, the predicted recovery rate is 10% which correspond to recover about 0.8M€. \n",
    "\n",
    "Class B conatins the NPLs which are likely to recover (Prob within 50% and 75%). A total of 135727 NPLs (7.2%) are in this class. The total debt corresponding to this class is 8.6M€, the predicted recovery rate is 10% which correspond to recover about 0.8M€. \n",
    "\n",
    "Class C contains the NPLs which are not likely tro recover (Prob within 25% and 50%). A total of 3847 NPLs (20%) are in this class. The total debt corresponding to this class is 25.8M€, the predicted recovery rate is about 7.5% which correspond to recover about 1.9M€. \n",
    "\n",
    "Finally class D contains the NPLs which are very unlikely to recover (Prob < 25%). A total of 12153 NPLs (64%) are in this class. The total debt corresponding to this class is about 72M€, the predicted recovery rate is 3.5% which correspond to recover about 2.5M€.\n",
    "\n",
    "Given the NPL sample as of 09-2017, the expected recovery rate that is between 1.6M€ (considering only A and B classes) and 6M€ (considering also C and D classes)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1700ab7f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "titles=['Number of customers', \n",
    "        'Fraction of customers', \n",
    "        'Predicted recovery rate',\n",
    "        'Debt gross book value (M€)', \n",
    "        'Predicted return (M€)']\n",
    "\n",
    "tmp=summary.copy()\n",
    "\n",
    "cols=['NUM_CUSTOMERS', 'FRAC_CUSTOMERS', 'PRED_TOTAL_RECOVERY_RATE', 'TOTAL_IMP_GBV', 'PRED_TOTAL_INCASSI', ]\n",
    "\n",
    "tmp[['TOTAL_IMP_GBV', 'PRED_TOTAL_INCASSI']] = tmp[['TOTAL_IMP_GBV', 'PRED_TOTAL_INCASSI']] / 1e6\n",
    "\n",
    "axs=tmp[cols].plot.bar(subplots=True, rot=0, figsize=(15,7),layout=(2,3), legend=False, title=titles)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}