52_xgb_dgi_base_label.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "4c65f557-651f-4958-a567-a1b6b5a82b2b",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Modify the file 00_setup.py to define input/output file paths on your system\n",
    "# The information in 00_setup.py will be used across notebooks\n",
    "from importlib.machinery import SourceFileLoader\n",
    "setup = SourceFileLoader(\"setup\", \"./00_setup.py\").load_module()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b01d0fdb-d4bf-465a-8541-997addc50b0c",
   "metadata": {},
   "source": [
    "# XGBoost, NAICS Embeddings from Unsupervised GNN (including target as feature)\n",
    "Retry an XGBoost model but use the NAICS embeddings generated by an unsupervised GNN (base features plus label)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "56161f89-1f54-47f9-860d-3f945a941c1f",
   "metadata": {},
   "source": [
    "*This script takes about 5 minutes on my MacBook Air*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "ec6b1778-7abc-4129-ab5f-ac2992bd8ca6",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import xgboost as xgb\n",
    "import warnings\n",
    "import matplotlib.pyplot as plt\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "785af927-a634-4e34-84c4-7a532236f2ca",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "fd09527b-32dd-47fa-b62f-f9307f89f553",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sklearn as sk\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "from sklearn.inspection import permutation_importance, partial_dependence\n",
    "from scikitplot.metrics import plot_lift_curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "0732ba40-8318-457b-b29f-f9b6de3b70aa",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sba_gnn.sba_gnn import sg_plot "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d14a5947-6599-4253-acf3-91ddeffdedd3",
   "metadata": {},
   "source": [
    "## Import Processed Datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f6bc5cc5-b635-420f-852c-6cc9bba68a52",
   "metadata": {},
   "outputs": [],
   "source": [
    "sba_loans = pd.read_parquet(Path(setup.temp_path).joinpath('01_DATA_transformed.parquet'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ad9ec376-52ba-4588-a35b-fa2b64bd4ce9",
   "metadata": {},
   "outputs": [],
   "source": [
    "embed_df = pd.read_parquet(Path(setup.temp_path).joinpath('33_DATA_embeddings.parquet')) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "4835b8e4-6c93-40a2-a713-3217c171e51f",
   "metadata": {},
   "outputs": [],
   "source": [
    "sba_loans = sba_loans.merge(embed_df.rename(columns={'NAICS_orig':'NAICS'}),\n",
    "                            on='NAICS', how='left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "a5f6c362-c1ab-48d8-812b-2b970b06586b",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = sba_loans[sba_loans['dset'] == 'train'].copy()\n",
    "test_df = sba_loans[sba_loans['dset'] == 'test']\n",
    "val_df = sba_loans[sba_loans['dset'] == 'val']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7b1db53b-181c-450d-9f7e-be350bc3111f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(436120, 64)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a6d13f46-cfc4-4351-a2a9-89fa42d34999",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(436120, 64)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.drop_duplicates('LoanNr_ChkDgt').shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d56e9533-8071-494a-a7da-4917a03fb701",
   "metadata": {},
   "source": [
    "#### Set the predictor features.  Base features and DGI embeddings (DGI features include label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c07ddd4e-6987-454a-926a-3b59abf6caa6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['NoEmp', 'CreateJob', 'LowDoc', 'DisbursementGross', 'new_business', 'urban_flag', 'franchise_flag', 'emb_000', 'emb_001', 'emb_002', 'emb_003', 'emb_004', 'emb_005', 'emb_006', 'emb_007']\n"
     ]
    }
   ],
   "source": [
    "predictor_features = setup.predictor_features + [c for c in embed_df.columns if c.startswith('emb')]\n",
    "print(predictor_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "51a66d4e-7977-4c8d-847e-e8fc0b5d8923",
   "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>NoEmp</th>\n",
       "      <th>CreateJob</th>\n",
       "      <th>LowDoc</th>\n",
       "      <th>DisbursementGross</th>\n",
       "      <th>new_business</th>\n",
       "      <th>urban_flag</th>\n",
       "      <th>franchise_flag</th>\n",
       "      <th>emb_000</th>\n",
       "      <th>emb_001</th>\n",
       "      <th>emb_002</th>\n",
       "      <th>emb_003</th>\n",
       "      <th>emb_004</th>\n",
       "      <th>emb_005</th>\n",
       "      <th>emb_006</th>\n",
       "      <th>emb_007</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>432893.000000</td>\n",
       "      <td>4.361200e+05</td>\n",
       "      <td>435584.000000</td>\n",
       "      <td>351624.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "      <td>436120.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>9.723145</td>\n",
       "      <td>2.164835</td>\n",
       "      <td>0.097777</td>\n",
       "      <td>1.944230e+05</td>\n",
       "      <td>0.281016</td>\n",
       "      <td>0.815075</td>\n",
       "      <td>0.052733</td>\n",
       "      <td>-0.262092</td>\n",
       "      <td>0.232365</td>\n",
       "      <td>0.239976</td>\n",
       "      <td>0.280361</td>\n",
       "      <td>0.146532</td>\n",
       "      <td>0.473102</td>\n",
       "      <td>-0.218652</td>\n",
       "      <td>0.308910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>61.690596</td>\n",
       "      <td>11.315009</td>\n",
       "      <td>0.297013</td>\n",
       "      <td>2.989452e+05</td>\n",
       "      <td>0.449496</td>\n",
       "      <td>0.388237</td>\n",
       "      <td>0.223501</td>\n",
       "      <td>0.019860</td>\n",
       "      <td>0.017607</td>\n",
       "      <td>0.018184</td>\n",
       "      <td>0.021244</td>\n",
       "      <td>0.348568</td>\n",
       "      <td>0.036070</td>\n",
       "      <td>0.397414</td>\n",
       "      <td>0.265269</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.000000e+03</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.312026</td>\n",
       "      <td>0.201386</td>\n",
       "      <td>0.207982</td>\n",
       "      <td>0.242983</td>\n",
       "      <td>-0.507427</td>\n",
       "      <td>0.342465</td>\n",
       "      <td>-0.606768</td>\n",
       "      <td>-0.474118</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>2.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3.572600e+04</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.279242</td>\n",
       "      <td>0.215951</td>\n",
       "      <td>0.223025</td>\n",
       "      <td>0.260557</td>\n",
       "      <td>-0.207332</td>\n",
       "      <td>0.442587</td>\n",
       "      <td>-0.534481</td>\n",
       "      <td>0.199513</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>4.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>9.000000e+04</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.261836</td>\n",
       "      <td>0.232138</td>\n",
       "      <td>0.239742</td>\n",
       "      <td>0.280087</td>\n",
       "      <td>0.305728</td>\n",
       "      <td>0.469931</td>\n",
       "      <td>-0.476766</td>\n",
       "      <td>0.428223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>9.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.148528e+05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-0.243578</td>\n",
       "      <td>0.247570</td>\n",
       "      <td>0.255679</td>\n",
       "      <td>0.298707</td>\n",
       "      <td>0.455578</td>\n",
       "      <td>0.500416</td>\n",
       "      <td>0.198722</td>\n",
       "      <td>0.473111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>9999.000000</td>\n",
       "      <td>3000.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.144632e+07</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.227149</td>\n",
       "      <td>0.276636</td>\n",
       "      <td>0.285697</td>\n",
       "      <td>0.333776</td>\n",
       "      <td>0.603074</td>\n",
       "      <td>0.550448</td>\n",
       "      <td>0.519361</td>\n",
       "      <td>0.604664</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               NoEmp      CreateJob         LowDoc  DisbursementGross  \\\n",
       "count  436120.000000  436120.000000  432893.000000       4.361200e+05   \n",
       "mean        9.723145       2.164835       0.097777       1.944230e+05   \n",
       "std        61.690596      11.315009       0.297013       2.989452e+05   \n",
       "min         0.000000       0.000000       0.000000       4.000000e+03   \n",
       "25%         2.000000       0.000000       0.000000       3.572600e+04   \n",
       "50%         4.000000       0.000000       0.000000       9.000000e+04   \n",
       "75%         9.000000       2.000000       0.000000       2.148528e+05   \n",
       "max      9999.000000    3000.000000       1.000000       1.144632e+07   \n",
       "\n",
       "        new_business     urban_flag  franchise_flag        emb_000  \\\n",
       "count  435584.000000  351624.000000   436120.000000  436120.000000   \n",
       "mean        0.281016       0.815075        0.052733      -0.262092   \n",
       "std         0.449496       0.388237        0.223501       0.019860   \n",
       "min         0.000000       0.000000        0.000000      -0.312026   \n",
       "25%         0.000000       1.000000        0.000000      -0.279242   \n",
       "50%         0.000000       1.000000        0.000000      -0.261836   \n",
       "75%         1.000000       1.000000        0.000000      -0.243578   \n",
       "max         1.000000       1.000000        1.000000      -0.227149   \n",
       "\n",
       "             emb_001        emb_002        emb_003        emb_004  \\\n",
       "count  436120.000000  436120.000000  436120.000000  436120.000000   \n",
       "mean        0.232365       0.239976       0.280361       0.146532   \n",
       "std         0.017607       0.018184       0.021244       0.348568   \n",
       "min         0.201386       0.207982       0.242983      -0.507427   \n",
       "25%         0.215951       0.223025       0.260557      -0.207332   \n",
       "50%         0.232138       0.239742       0.280087       0.305728   \n",
       "75%         0.247570       0.255679       0.298707       0.455578   \n",
       "max         0.276636       0.285697       0.333776       0.603074   \n",
       "\n",
       "             emb_005        emb_006        emb_007  \n",
       "count  436120.000000  436120.000000  436120.000000  \n",
       "mean        0.473102      -0.218652       0.308910  \n",
       "std         0.036070       0.397414       0.265269  \n",
       "min         0.342465      -0.606768      -0.474118  \n",
       "25%         0.442587      -0.534481       0.199513  \n",
       "50%         0.469931      -0.476766       0.428223  \n",
       "75%         0.500416       0.198722       0.473111  \n",
       "max         0.550448       0.519361       0.604664  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df[predictor_features].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2a4f4a9e-f7e5-4143-b1e4-b5fbb5f7280d",
   "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>NoEmp</th>\n",
       "      <th>CreateJob</th>\n",
       "      <th>LowDoc</th>\n",
       "      <th>DisbursementGross</th>\n",
       "      <th>new_business</th>\n",
       "      <th>urban_flag</th>\n",
       "      <th>franchise_flag</th>\n",
       "      <th>emb_000</th>\n",
       "      <th>emb_001</th>\n",
       "      <th>emb_002</th>\n",
       "      <th>emb_003</th>\n",
       "      <th>emb_004</th>\n",
       "      <th>emb_005</th>\n",
       "      <th>emb_006</th>\n",
       "      <th>emb_007</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <td>-0.027403</td>\n",
       "      <td>-0.010002</td>\n",
       "      <td>-0.079523</td>\n",
       "      <td>-0.120788</td>\n",
       "      <td>0.013878</td>\n",
       "      <td>0.053785</td>\n",
       "      <td>-0.012731</td>\n",
       "      <td>-0.031886</td>\n",
       "      <td>0.031886</td>\n",
       "      <td>0.031886</td>\n",
       "      <td>0.031886</td>\n",
       "      <td>-0.160686</td>\n",
       "      <td>0.035261</td>\n",
       "      <td>-0.14905</td>\n",
       "      <td>-0.016147</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           NoEmp  CreateJob    LowDoc  DisbursementGross  new_business  \\\n",
       "target -0.027403  -0.010002 -0.079523          -0.120788      0.013878   \n",
       "\n",
       "        urban_flag  franchise_flag   emb_000   emb_001   emb_002   emb_003  \\\n",
       "target    0.053785       -0.012731 -0.031886  0.031886  0.031886  0.031886   \n",
       "\n",
       "         emb_004   emb_005  emb_006   emb_007  target  \n",
       "target -0.160686  0.035261 -0.14905 -0.016147     1.0  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tcorr = train_df[predictor_features + ['target']].corr()\n",
    "tcorr.to_csv(Path(setup.temp_path).joinpath('52_REPORT_correlations.csv'))\n",
    "tcorr[tcorr.index == 'target']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "1f8c475d-7ead-4f92-a19b-326343528f59",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(Path(setup.temp_path).joinpath('52_DATA_features_predict.pkl'), 'wb') as fout:\n",
    "    pickle.dump(predictor_features, fout)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b204a92-d109-4f6d-87a5-3d5dc05d2e5f",
   "metadata": {},
   "source": [
    "## Model Fit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "af48634b-fc39-4c29-9792-bcf1cdf241c0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'subsample': 0.8,\n",
       " 'scale_pos_weight': 1.0,\n",
       " 'reg_lambda': 0.01,\n",
       " 'reg_alpha': 0.001,\n",
       " 'min_child_weight': 50,\n",
       " 'max_depth': 8,\n",
       " 'learning_rate': 0.1,\n",
       " 'gamma': 2}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_params_df = pd.read_csv(Path(setup.temp_path).joinpath('03_REPORT_fit_parameter_selected.csv'))\n",
    "best_params = best_params_df.to_dict(orient='records')[0]\n",
    "best_params"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bc698909-3663-440f-98ce-fcfe312faa82",
   "metadata": {},
   "source": [
    "##### Fit using pre-determined parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "e8eca4a6-66e8-4eb2-b68a-1271f59cdbec",
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_params = dict(best_params, \n",
    "                   **{'objective':\"binary:logistic\", 'n_estimators':setup.xgb_n_estimators})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "6d124de6-1d5d-4ff9-8071-4d05b0717f13",
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_model = xgb.XGBClassifier(objective=\"binary:logistic\",\n",
    "                             n_estimators = setup.xgb_n_estimators)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "9eebd329-7e39-4e52-be94-d839ad559b5d",
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_model = xgb_model.set_params(**xgb_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "a7f80910-9719-486b-8eac-93889064708a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-logloss:0.50158\tvalidation_1-logloss:0.50502\n",
      "[1]\tvalidation_0-logloss:0.49498\tvalidation_1-logloss:0.49856\n",
      "[2]\tvalidation_0-logloss:0.48946\tvalidation_1-logloss:0.49313\n",
      "[3]\tvalidation_0-logloss:0.48456\tvalidation_1-logloss:0.48830\n",
      "[4]\tvalidation_0-logloss:0.48040\tvalidation_1-logloss:0.48429\n",
      "[5]\tvalidation_0-logloss:0.47682\tvalidation_1-logloss:0.48078\n",
      "[6]\tvalidation_0-logloss:0.47374\tvalidation_1-logloss:0.47780\n",
      "[7]\tvalidation_0-logloss:0.47107\tvalidation_1-logloss:0.47519\n",
      "[8]\tvalidation_0-logloss:0.46878\tvalidation_1-logloss:0.47301\n",
      "[9]\tvalidation_0-logloss:0.46678\tvalidation_1-logloss:0.47114\n",
      "[10]\tvalidation_0-logloss:0.46500\tvalidation_1-logloss:0.46939\n",
      "[11]\tvalidation_0-logloss:0.46353\tvalidation_1-logloss:0.46800\n",
      "[12]\tvalidation_0-logloss:0.46206\tvalidation_1-logloss:0.46662\n",
      "[13]\tvalidation_0-logloss:0.46078\tvalidation_1-logloss:0.46539\n",
      "[14]\tvalidation_0-logloss:0.45972\tvalidation_1-logloss:0.46443\n",
      "[15]\tvalidation_0-logloss:0.45880\tvalidation_1-logloss:0.46360\n",
      "[16]\tvalidation_0-logloss:0.45786\tvalidation_1-logloss:0.46275\n",
      "[17]\tvalidation_0-logloss:0.45709\tvalidation_1-logloss:0.46206\n",
      "[18]\tvalidation_0-logloss:0.45643\tvalidation_1-logloss:0.46148\n",
      "[19]\tvalidation_0-logloss:0.45581\tvalidation_1-logloss:0.46095\n",
      "[20]\tvalidation_0-logloss:0.45523\tvalidation_1-logloss:0.46046\n",
      "[21]\tvalidation_0-logloss:0.45467\tvalidation_1-logloss:0.45999\n",
      "[22]\tvalidation_0-logloss:0.45420\tvalidation_1-logloss:0.45962\n",
      "[23]\tvalidation_0-logloss:0.45373\tvalidation_1-logloss:0.45923\n",
      "[24]\tvalidation_0-logloss:0.45335\tvalidation_1-logloss:0.45896\n",
      "[25]\tvalidation_0-logloss:0.45299\tvalidation_1-logloss:0.45868\n",
      "[26]\tvalidation_0-logloss:0.45266\tvalidation_1-logloss:0.45844\n",
      "[27]\tvalidation_0-logloss:0.45238\tvalidation_1-logloss:0.45818\n",
      "[28]\tvalidation_0-logloss:0.45203\tvalidation_1-logloss:0.45790\n",
      "[29]\tvalidation_0-logloss:0.45167\tvalidation_1-logloss:0.45760\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-1 {color: black;}#sk-container-id-1 pre{padding: 0;}#sk-container-id-1 div.sk-toggleable {background-color: white;}#sk-container-id-1 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-1 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-1 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-1 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-1 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-1 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-1 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-1 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-1 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-1 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-1 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-1 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-1 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-1 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-1 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-1 div.sk-item {position: relative;z-index: 1;}#sk-container-id-1 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-1 div.sk-item::before, #sk-container-id-1 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-1 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-1 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-1 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-1 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-1 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-1 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-1 div.sk-label-container {text-align: center;}#sk-container-id-1 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-1 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-1\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>XGBClassifier(base_score=None, booster=None, callbacks=None,\n",
       "              colsample_bylevel=None, colsample_bynode=None,\n",
       "              colsample_bytree=None, device=None, early_stopping_rounds=None,\n",
       "              enable_categorical=False, eval_metric=None, feature_types=None,\n",
       "              gamma=2, grow_policy=None, importance_type=None,\n",
       "              interaction_constraints=None, learning_rate=0.1, max_bin=None,\n",
       "              max_cat_threshold=None, max_cat_to_onehot=None,\n",
       "              max_delta_step=None, max_depth=8, max_leaves=None,\n",
       "              min_child_weight=50, missing=nan, monotone_constraints=None,\n",
       "              multi_strategy=None, n_estimators=30, n_jobs=None,\n",
       "              num_parallel_tree=None, random_state=None, ...)</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-1\" type=\"checkbox\" checked><label for=\"sk-estimator-id-1\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">XGBClassifier</label><div class=\"sk-toggleable__content\"><pre>XGBClassifier(base_score=None, booster=None, callbacks=None,\n",
       "              colsample_bylevel=None, colsample_bynode=None,\n",
       "              colsample_bytree=None, device=None, early_stopping_rounds=None,\n",
       "              enable_categorical=False, eval_metric=None, feature_types=None,\n",
       "              gamma=2, grow_policy=None, importance_type=None,\n",
       "              interaction_constraints=None, learning_rate=0.1, max_bin=None,\n",
       "              max_cat_threshold=None, max_cat_to_onehot=None,\n",
       "              max_delta_step=None, max_depth=8, max_leaves=None,\n",
       "              min_child_weight=50, missing=nan, monotone_constraints=None,\n",
       "              multi_strategy=None, n_estimators=30, n_jobs=None,\n",
       "              num_parallel_tree=None, random_state=None, ...)</pre></div></div></div></div></div>"
      ],
      "text/plain": [
       "XGBClassifier(base_score=None, booster=None, callbacks=None,\n",
       "              colsample_bylevel=None, colsample_bynode=None,\n",
       "              colsample_bytree=None, device=None, early_stopping_rounds=None,\n",
       "              enable_categorical=False, eval_metric=None, feature_types=None,\n",
       "              gamma=2, grow_policy=None, importance_type=None,\n",
       "              interaction_constraints=None, learning_rate=0.1, max_bin=None,\n",
       "              max_cat_threshold=None, max_cat_to_onehot=None,\n",
       "              max_delta_step=None, max_depth=8, max_leaves=None,\n",
       "              min_child_weight=50, missing=nan, monotone_constraints=None,\n",
       "              multi_strategy=None, n_estimators=30, n_jobs=None,\n",
       "              num_parallel_tree=None, random_state=None, ...)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb_model.fit(train_df[predictor_features], train_df['target'],\n",
    "              eval_set = [(train_df[predictor_features], train_df['target']),\n",
    "                         (val_df[predictor_features], val_df['target'])])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0913a102-a730-4b94-9977-e584df764d64",
   "metadata": {},
   "source": [
    "##### Save the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "af21dd7c-09d4-407f-9960-0748ba8f2247",
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb_model.save_model(Path(setup.temp_path).joinpath('52_MODEL_xgboost.json'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8e5d3e7d-0bd5-4252-b287-c9cc65f7a132",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(Path(setup.temp_path).joinpath('52_MODEL_xgboost.pkl'), 'wb') as po:\n",
    "    pickle.dump(xgb_model, po)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f34ad059-70dc-4672-adff-0d961fc74b07",
   "metadata": {},
   "source": [
    "## Probability Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "6a7c8509-0109-4f8e-a466-d6e21164012c",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_df = pd.concat([sba_loans[['LoanNr_ChkDgt', 'dset', 'target', 'dset_naics_holdout']].reset_index(),\n",
    "                     pd.DataFrame({'predict_prob':\n",
    "                                   xgb_model.predict_proba(sba_loans[predictor_features])[:,1]})],\n",
    "                    axis=1) \\\n",
    "    .set_index('index')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "7084012d-b68a-4ca4-b714-160f8d564500",
   "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>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>max</th>\n",
       "      <th>median</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</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>0</th>\n",
       "      <td>0.186632</td>\n",
       "      <td>0.109991</td>\n",
       "      <td>0.014645</td>\n",
       "      <td>0.662964</td>\n",
       "      <td>0.180411</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.277651</td>\n",
       "      <td>0.098221</td>\n",
       "      <td>0.015077</td>\n",
       "      <td>0.678232</td>\n",
       "      <td>0.287210</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            mean       std       min       max    median\n",
       "target                                                  \n",
       "0       0.186632  0.109991  0.014645  0.662964  0.180411\n",
       "1       0.277651  0.098221  0.015077  0.678232  0.287210"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df.groupby('target')['predict_prob'].agg(['mean', 'std', 'min', 'max', 'median'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d33cb4dc-6c89-4a37-98c7-d3a666abc4a5",
   "metadata": {},
   "source": [
    "## Decision Threshold\n",
    "Max f1, on training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "d7727498-825f-4f53-882f-cfcce730b1fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_train_df = pred_df[pred_df['dset'] == 'train']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "7b553798-eead-48f8-b14b-5f036d9183d9",
   "metadata": {},
   "outputs": [],
   "source": [
    "thresh_tune_data = sg_plot.get_f1_frame(pred_train_df['target'], \n",
    "                                        pred_train_df['predict_prob'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "7a1cde01-a30b-475f-8fd7-409b13946c6d",
   "metadata": {},
   "outputs": [],
   "source": [
    "thresh_tune_data.to_csv(Path(setup.temp_path).joinpath('52_REPORT_thresh_tune.csv'), index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "947a487c-e71d-433c-bbf0-2ed7e57a908a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.24\n"
     ]
    },
    {
     "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>thresh</th>\n",
       "      <th>f1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.24</td>\n",
       "      <td>0.450514</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.22</td>\n",
       "      <td>0.449453</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    thresh        f1\n",
       "12    0.24  0.450514\n",
       "11    0.22  0.449453"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thresh_head = thresh_tune_data.sort_values('f1', ascending=False).head(2)\n",
    "best_thresh = thresh_head['thresh'].iloc[0]\n",
    "print(best_thresh)\n",
    "thresh_head"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ee5330e-c1a7-4310-afbe-be9bc0d26f43",
   "metadata": {},
   "source": [
    "##### Apply decision threshold\n",
    "All datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "f0d4c1cf-6d5b-49e9-9123-a067697a9e39",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_df['predict_bin'] = np.where(pred_df['predict_prob'] >= best_thresh, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "8253c32c-0438-47b9-8570-7cde6f7f5590",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "predict_bin\n",
       "0    398911\n",
       "1    289170\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_df['predict_bin'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "dbb46f4a-a5ef-4fd6-9da4-30de4a00334d",
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_df.to_parquet(Path(setup.temp_path).joinpath('52_DATA_combined_predictions.parquet'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ab860d0-eddc-4f63-b853-88da15d451f4",
   "metadata": {},
   "source": [
    "## Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "f2801964-124e-425f-b4cb-9bcf1cf3f04a",
   "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>dset</th>\n",
       "      <th>accuracy_score</th>\n",
       "      <th>f1_score</th>\n",
       "      <th>precision_score</th>\n",
       "      <th>recall_score</th>\n",
       "      <th>average_precision_score</th>\n",
       "      <th>roc_auc_score</th>\n",
       "      <th>dset_naics_holdout</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.660425</td>\n",
       "      <td>0.439831</td>\n",
       "      <td>0.324877</td>\n",
       "      <td>0.680679</td>\n",
       "      <td>0.366782</td>\n",
       "      <td>0.729274</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train</td>\n",
       "      <td>0.654996</td>\n",
       "      <td>0.450514</td>\n",
       "      <td>0.334186</td>\n",
       "      <td>0.691073</td>\n",
       "      <td>0.382039</td>\n",
       "      <td>0.729314</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val</td>\n",
       "      <td>0.652471</td>\n",
       "      <td>0.449021</td>\n",
       "      <td>0.334564</td>\n",
       "      <td>0.682517</td>\n",
       "      <td>0.378050</td>\n",
       "      <td>0.723115</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.651319</td>\n",
       "      <td>0.446626</td>\n",
       "      <td>0.331543</td>\n",
       "      <td>0.684076</td>\n",
       "      <td>0.373773</td>\n",
       "      <td>0.723915</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>test</td>\n",
       "      <td>0.673507</td>\n",
       "      <td>0.429074</td>\n",
       "      <td>0.314460</td>\n",
       "      <td>0.675154</td>\n",
       "      <td>0.355020</td>\n",
       "      <td>0.735716</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    dset  accuracy_score  f1_score  precision_score  recall_score  \\\n",
       "0   test        0.660425  0.439831         0.324877      0.680679   \n",
       "1  train        0.654996  0.450514         0.334186      0.691073   \n",
       "2    val        0.652471  0.449021         0.334564      0.682517   \n",
       "0   test        0.651319  0.446626         0.331543      0.684076   \n",
       "1   test        0.673507  0.429074         0.314460      0.675154   \n",
       "\n",
       "   average_precision_score  roc_auc_score  dset_naics_holdout  \n",
       "0                 0.366782       0.729274                 NaN  \n",
       "1                 0.382039       0.729314                 NaN  \n",
       "2                 0.378050       0.723115                 NaN  \n",
       "0                 0.373773       0.723915                 0.0  \n",
       "1                 0.355020       0.735716                 1.0  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics_dset_df = pred_df.groupby('dset') \\\n",
    "    .apply(lambda x: sg_plot.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_test_df = pred_df[pred_df['dset'] == 'test'] \\\n",
    "    .groupby(['dset', 'dset_naics_holdout']) \\\n",
    "    .apply(lambda x: sg_plot.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_df = pd.concat([metrics_dset_df, metrics_test_df])\n",
    "metrics_df.to_csv(Path(setup.temp_path).joinpath('52_REPORT_metrics.csv'), index=True)\n",
    "metrics_df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d906270-4264-4b0b-9f50-dcef8f75a5fd",
   "metadata": {},
   "source": [
    "## Lift\n",
    "Test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "82d9c0ab-8989-450d-a2b1-7997ca636052",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "pred_test_df = pred_df[pred_df['dset'] == 'test']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "c01f5279-1151-4a8d-9b21-e12d4092f3c1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_lift_curve(pred_test_df['target'], \n",
    "                pd.concat([1-pred_test_df['predict_prob'], pred_test_df['predict_prob']], axis=1), \n",
    "                title='Lift Curve', ax=None, figsize=(4,3), \n",
    "                title_fontsize='large', text_fontsize='medium')\n",
    "plt.gca().legend(loc='upper right')\n",
    "plt.gcf().savefig(Path(setup.temp_path).joinpath('52_PLOT_lift_curve.png'),bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "04994f41-3c98-412a-bb2f-11e5ae22b432",
   "metadata": {},
   "source": [
    "## Feature Importance "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "926af7df-e407-484b-bd86-ae7d76a46392",
   "metadata": {},
   "source": [
    "##### XGBoost gain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "ec3f6e1f-2b29-4b46-ac12-7a8451a3a45b",
   "metadata": {},
   "outputs": [],
   "source": [
    "feat_impgain = pd.DataFrame(xgb_model.get_booster().get_score(importance_type='gain'),\n",
    "                                     index=['importance']).T \\\n",
    "    .sort_values('importance', ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "9586c637-7add-4401-b61c-37abd55b11b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "feat_impgain.to_csv(Path(setup.temp_path).joinpath('52_REPORT_importance_gain.csv'), index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "311e111f-d8e5-4a52-8adc-6a7c5f9daa8f",
   "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>importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>urban_flag</th>\n",
       "      <td>879.199036</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LowDoc</th>\n",
       "      <td>196.701263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>DisbursementGross</th>\n",
       "      <td>100.682716</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>emb_004</th>\n",
       "      <td>42.749302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>franchise_flag</th>\n",
       "      <td>31.256042</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   importance\n",
       "urban_flag         879.199036\n",
       "LowDoc             196.701263\n",
       "DisbursementGross  100.682716\n",
       "emb_004             42.749302\n",
       "franchise_flag      31.256042"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feat_impgain.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "9c88564a-7289-4c49-b2cd-b97338d1b042",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "feat_impgain.plot(kind='barh', legend=None)\n",
    "plt.gca().invert_yaxis()\n",
    "plt.xlabel('gain importance')\n",
    "plt.gcf().savefig(Path(setup.temp_path).joinpath('52_PLOT_importance_gain.png'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "27e5fb90-f305-40ff-acaf-285d7d056c15",
   "metadata": {},
   "source": [
    "##### Permutation Importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "e2c1c123-20d0-423d-b684-585c0a5d6d2b",
   "metadata": {},
   "outputs": [],
   "source": [
    "perm_importance = permutation_importance(xgb_model, \n",
    "                                         train_df[predictor_features],  train_df['target'],\n",
    "                                         n_repeats=30, scoring = 'average_precision')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "ab0df42e-af4e-46b1-b0cf-630877137678",
   "metadata": {},
   "outputs": [],
   "source": [
    "perm_importance_df = pd.DataFrame({'features':predictor_features,\n",
    "                                   'importance': perm_importance['importances_mean']}) \n",
    "perm_importance_df['abs_importance'] = np.abs(perm_importance_df['importance'])\n",
    "perm_importance_df = perm_importance_df.sort_values('abs_importance', ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "0cbc5d84-6775-404f-a685-44f5be5cd481",
   "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>features</th>\n",
       "      <th>importance</th>\n",
       "      <th>abs_importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>DisbursementGross</td>\n",
       "      <td>0.076575</td>\n",
       "      <td>0.076575</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>urban_flag</td>\n",
       "      <td>0.038891</td>\n",
       "      <td>0.038891</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>emb_004</td>\n",
       "      <td>0.033459</td>\n",
       "      <td>0.033459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>CreateJob</td>\n",
       "      <td>0.026448</td>\n",
       "      <td>0.026448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>LowDoc</td>\n",
       "      <td>0.015590</td>\n",
       "      <td>0.015590</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             features  importance  abs_importance\n",
       "3   DisbursementGross    0.076575        0.076575\n",
       "5          urban_flag    0.038891        0.038891\n",
       "11            emb_004    0.033459        0.033459\n",
       "1           CreateJob    0.026448        0.026448\n",
       "2              LowDoc    0.015590        0.015590"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perm_importance_df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "38059608-2f72-491f-86b3-0c1418b83531",
   "metadata": {},
   "outputs": [],
   "source": [
    "perm_importance_df \\\n",
    "    .to_csv(Path(setup.temp_path).joinpath('52_DATA_importance_perturb.csv'),\n",
    "            index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "c27499c8-065b-4143-a7dc-bce301bf3edf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "perm_importance_df.head(10).plot(kind='barh', x='features', y='importance', legend=None)\n",
    "plt.gca().invert_yaxis()\n",
    "plt.xlabel('permutation importance')\n",
    "plt.ylabel(None)\n",
    "plt.gcf().savefig(Path(setup.temp_path).joinpath('52_PLOT_importance_perturb.png'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6fd44a7c-bcf5-4284-8b9f-b575fa523ff3",
   "metadata": {},
   "source": [
    "## PDP Plot - Embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "1a29e744-7088-41ca-9ed8-48f0c4d21a45",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_pdp(feature):\n",
    "    results = partial_dependence(xgb_model, train_df[predictor_features], [feature])\n",
    "    return pd.DataFrame({'grid_values':results[\"grid_values\"][0],\n",
    "                       'pdp':results[\"average\"][0]})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e19612cd-2f49-469c-9ac1-80228c8a7b97",
   "metadata": {},
   "outputs": [],
   "source": [
    "emb_feat = [c for c in embed_df.columns if c.startswith('emb')]\n",
    "pdp_res = pd.concat([get_pdp(c) for c in emb_feat], axis=0, keys=emb_feat) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "81abd433-fe66-41d6-a1cf-774b751e6555",
   "metadata": {},
   "outputs": [],
   "source": [
    "pdp_res = pdp_res.reset_index(level=0) \\\n",
    "    .rename(columns={'level_0':'feature'})\n",
    "pdp_res.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ab10a220-41e8-4e78-bf17-12febc5937ca",
   "metadata": {},
   "outputs": [],
   "source": [
    "nrow = int(np.ceil(len(emb_feat) / 2))\n",
    "ncol = 2 if len(emb_feat) % 2 == 0 else 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1619911c-f318-4202-b3c4-6eb62f2b9e2b",
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots(nrow, ncol, figsize = (10, 8), sharex = False, sharey=True)\n",
    "for i in range(len(emb_feat)):\n",
    "    this_row = int(np.floor(i / 2))\n",
    "    this_col = int(i % 2)\n",
    "    this_pdp = pdp_res[pdp_res['feature'] == emb_feat[i]]\n",
    "    #print(f'i:{i}, row:{this_row}, col:{this_col}, feat:{emb_feat[i]}')\n",
    "    this_pdp.plot(x='grid_values', y='pdp', label=emb_feat[i], ax=ax[this_row, this_col])\n",
    "    #ax[this_row, this_col].set_title(emb_feat[i])\n",
    "    #ax[this_row, this_col].legend(loc='upper left')\n",
    "    ax[this_row, this_col].set_xlabel(None)\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "fig.savefig(Path(setup.temp_path).joinpath('52_PLOT_pdp_naics_embeddings.png'),\n",
    "               bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5488c384-feb9-45b4-8d95-e2ea4954719b",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a5053579-500f-40bb-abb2-70b55c041214",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cc45aede-50bb-47a4-90ca-2aecc989738a",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf_p39",
   "language": "python",
   "name": "tf_p39"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}