deep-learning-for-tabular-data.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
    "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
    "collapsed": true
   },
   "source": [
    "![](https://storage.googleapis.com/kaggle-competitions/kaggle/3338/media/gate.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_uuid": "b4c52628791988389da2992adf97768172b19eb8"
   },
   "source": [
    "## Amazon.com - Employee Access Challenge\n",
    "\n",
    "When an employee at any company starts work, they first need to obtain the computer access necessary to fulfill their role. This access may allow an employee to read/manipulate resources through various applications or web portals. It is assumed that employees fulfilling the functions of a given role will access the same or similar resources. It is often the case that employees figure out the access they need as they encounter roadblocks during their daily work (e.g. not able to log into a reporting portal). A knowledgeable supervisor then takes time to manually grant the needed access in order to overcome access obstacles. As employees move throughout a company, this access discovery/recovery cycle wastes a nontrivial amount of time and money.\n",
    "\n",
    "There is a considerable amount of data regarding an employee’s role within an organization and the resources to which they have access. Given the data related to current employees and their provisioned access, models can be built that automatically determine access privileges as employees enter and leave roles within a company. These auto-access models seek to minimize the human involvement required to grant or revoke employee access."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Kernel's motivations\n",
    "\n",
    "In this kernel we are going to show how deep learning (using TensorFlow 2.0 and Keras) can be effectively used when the problem involved regards tabular data. We will compare a deep neural network solution (DNN) to the best in class gradient boosting machine (GBM) algorithm when high cardinality variables are present and you will discover how not only a DNN solution is comparable, but also how it can integrate nicely with a GBM solution."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using a datagenerator for tabular data\n",
    "The process of preparing the data is made easier and more efficient using a datagenerator from https://github.com/lmassaron/deep_learning_for_tabular_data The generator can handle into different pipelines numeric, ordinal, categorical (bot low and high cardinality) variables and it automatically takes care of missing values. You can also find in the repository a couple of useful activations for tabular data models: gelu (https://arxiv.org/abs/1606.08415) and mish (https://arxiv.org/abs/1908.08681)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_uuid": "c092d3a33400d1d95e79c5c33ef9f7227ea491ab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting catboost\r\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/ca/ae/aaff63662f7f5d2af7ec8d61a6f39e78ada9348e5df4f43e665ecc4bea10/catboost-0.21-cp36-none-manylinux1_x86_64.whl (64.0MB)\r\n",
      "\u001b[K     |████████████████████████████████| 64.0MB 35.7MB/s \r\n",
      "\u001b[?25hRequirement already satisfied, skipping upgrade: graphviz in /opt/conda/lib/python3.6/site-packages (from catboost) (0.8.4)\r\n",
      "Requirement already satisfied, skipping upgrade: plotly in /opt/conda/lib/python3.6/site-packages (from catboost) (4.1.1)\r\n",
      "Requirement already satisfied, skipping upgrade: six in /opt/conda/lib/python3.6/site-packages (from catboost) (1.12.0)\r\n",
      "Requirement already satisfied, skipping upgrade: numpy>=1.16.0 in /opt/conda/lib/python3.6/site-packages (from catboost) (1.16.4)\r\n",
      "Requirement already satisfied, skipping upgrade: pandas>=0.24.0 in /opt/conda/lib/python3.6/site-packages (from catboost) (0.25.1)\r\n",
      "Requirement already satisfied, skipping upgrade: matplotlib in /opt/conda/lib/python3.6/site-packages (from catboost) (3.0.3)\r\n",
      "Requirement already satisfied, skipping upgrade: scipy in /opt/conda/lib/python3.6/site-packages (from catboost) (1.2.1)\r\n",
      "Requirement already satisfied, skipping upgrade: retrying>=1.3.3 in /opt/conda/lib/python3.6/site-packages (from plotly->catboost) (1.3.3)\r\n",
      "Requirement already satisfied, skipping upgrade: python-dateutil>=2.6.1 in /opt/conda/lib/python3.6/site-packages (from pandas>=0.24.0->catboost) (2.8.0)\r\n",
      "Requirement already satisfied, skipping upgrade: pytz>=2017.2 in /opt/conda/lib/python3.6/site-packages (from pandas>=0.24.0->catboost) (2019.2)\r\n",
      "Requirement already satisfied, skipping upgrade: cycler>=0.10 in /opt/conda/lib/python3.6/site-packages (from matplotlib->catboost) (0.10.0)\r\n",
      "Requirement already satisfied, skipping upgrade: kiwisolver>=1.0.1 in /opt/conda/lib/python3.6/site-packages (from matplotlib->catboost) (1.1.0)\r\n",
      "Requirement already satisfied, skipping upgrade: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /opt/conda/lib/python3.6/site-packages (from matplotlib->catboost) (2.4.2)\r\n",
      "Requirement already satisfied, skipping upgrade: setuptools in /opt/conda/lib/python3.6/site-packages (from kiwisolver>=1.0.1->matplotlib->catboost) (41.2.0)\r\n",
      "Installing collected packages: catboost\r\n",
      "  Found existing installation: catboost 0.17.3\r\n",
      "    Uninstalling catboost-0.17.3:\r\n",
      "      Successfully uninstalled catboost-0.17.3\r\n",
      "Successfully installed catboost-0.21\r\n"
     ]
    }
   ],
   "source": [
    "# Assuring you have the most recent CatBoost release\n",
    "!pip install catboost -U"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cloning into 'deep_learning_for_tabular_data'...\r\n",
      "remote: Enumerating objects: 34, done.\u001b[K\r\n",
      "remote: Counting objects: 100% (34/34), done.\u001b[K\r\n",
      "remote: Compressing objects: 100% (31/31), done.\u001b[K\r\n",
      "remote: Total 34 (delta 15), reused 5 (delta 1), pack-reused 0\u001b[K\r\n",
      "Unpacking objects: 100% (34/34), done.\r\n"
     ]
    }
   ],
   "source": [
    "# Getting useful tabular processing and generator functions\n",
    "!git clone https://github.com/lmassaron/deep_learning_for_tabular_data.git"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "_uuid": "5a24f34eb52a736e12dddfd2cecc3b6a996e6442"
   },
   "outputs": [],
   "source": [
    "# Importing core libraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from time import time\n",
    "import pprint\n",
    "import joblib\n",
    "\n",
    "# Suppressing warnings because of skopt verbosity\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "# Classifiers\n",
    "from catboost import CatBoostClassifier, Pool\n",
    "\n",
    "# Model selection\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "\n",
    "# Metrics\n",
    "from sklearn.metrics import roc_auc_score, average_precision_score\n",
    "from sklearn.metrics import make_scorer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "_uuid": "652842d2daf545724e6dcdae6fba1c3009d09d71"
   },
   "outputs": [],
   "source": [
    "# Loading data directly from CatBoost\n",
    "from catboost.datasets import amazon\n",
    "\n",
    "X, Xt = amazon()\n",
    "\n",
    "y = X[\"ACTION\"].apply(lambda x: 1 if x == 1 else 0).values\n",
    "X.drop([\"ACTION\"], axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Transforming all the labels of all variables\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "label_encoders = [LabelEncoder() for _ in range(X.shape[1])]\n",
    "\n",
    "for col, column in enumerate(X.columns):\n",
    "    label_encoders[col].fit(X[column].append(Xt[column]))\n",
    "    X[column] = label_encoders[col].transform(X[column])\n",
    "    Xt[column] = label_encoders[col].transform(Xt[column])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Enconding frequencies instead of labels (so we have some numeric variables)\n",
    "\n",
    "def frequency_encoding(column, df, df_test=None):\n",
    "    frequencies = df[column].value_counts().reset_index()\n",
    "    df_values = df[[column]].merge(frequencies, how='left', \n",
    "                                   left_on=column, right_on='index').iloc[:,-1].values\n",
    "    if df_test is not None:\n",
    "        df_test_values = df_test[[column]].merge(frequencies, how='left', \n",
    "                                                 left_on=column, right_on='index').fillna(1).iloc[:,-1].values\n",
    "    else:\n",
    "        df_test_values = None\n",
    "    return df_values, df_test_values\n",
    "\n",
    "for column in X.columns:\n",
    "    train_values, test_values = frequency_encoding(column, X, Xt)\n",
    "    X[column+'_counts'] = train_values\n",
    "    Xt[column+'_counts'] = test_values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Pointing out which variables are categorical and which are numeric\n",
    "categorical_variables = [col for col in X.columns if '_counts' not in col]\n",
    "numeric_variables = [col for col in X.columns if '_counts' in col]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "_uuid": "b856921d70bed3e3ed231d3ab76b3161915630b0"
   },
   "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>RESOURCE</th>\n",
       "      <th>MGR_ID</th>\n",
       "      <th>ROLE_ROLLUP_1</th>\n",
       "      <th>ROLE_ROLLUP_2</th>\n",
       "      <th>ROLE_DEPTNAME</th>\n",
       "      <th>ROLE_TITLE</th>\n",
       "      <th>ROLE_FAMILY_DESC</th>\n",
       "      <th>ROLE_FAMILY</th>\n",
       "      <th>ROLE_CODE</th>\n",
       "      <th>RESOURCE_counts</th>\n",
       "      <th>MGR_ID_counts</th>\n",
       "      <th>ROLE_ROLLUP_1_counts</th>\n",
       "      <th>ROLE_ROLLUP_2_counts</th>\n",
       "      <th>ROLE_DEPTNAME_counts</th>\n",
       "      <th>ROLE_TITLE_counts</th>\n",
       "      <th>ROLE_FAMILY_DESC_counts</th>\n",
       "      <th>ROLE_FAMILY_counts</th>\n",
       "      <th>ROLE_CODE_counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>3050</td>\n",
       "      <td>4440</td>\n",
       "      <td>21</td>\n",
       "      <td>65</td>\n",
       "      <td>319</td>\n",
       "      <td>4</td>\n",
       "      <td>7</td>\n",
       "      <td>65</td>\n",
       "      <td>4</td>\n",
       "      <td>3</td>\n",
       "      <td>55</td>\n",
       "      <td>21407</td>\n",
       "      <td>4424</td>\n",
       "      <td>72</td>\n",
       "      <td>3583</td>\n",
       "      <td>6896</td>\n",
       "      <td>10980</td>\n",
       "      <td>3583</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>644</td>\n",
       "      <td>162</td>\n",
       "      <td>21</td>\n",
       "      <td>69</td>\n",
       "      <td>310</td>\n",
       "      <td>34</td>\n",
       "      <td>62</td>\n",
       "      <td>67</td>\n",
       "      <td>38</td>\n",
       "      <td>30</td>\n",
       "      <td>10</td>\n",
       "      <td>21407</td>\n",
       "      <td>3945</td>\n",
       "      <td>159</td>\n",
       "      <td>81</td>\n",
       "      <td>12</td>\n",
       "      <td>1287</td>\n",
       "      <td>81</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2706</td>\n",
       "      <td>1679</td>\n",
       "      <td>50</td>\n",
       "      <td>58</td>\n",
       "      <td>14</td>\n",
       "      <td>0</td>\n",
       "      <td>2590</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>184</td>\n",
       "      <td>184</td>\n",
       "      <td>546</td>\n",
       "      <td>1256</td>\n",
       "      <td>33</td>\n",
       "      <td>2636</td>\n",
       "      <td>1256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>2615</td>\n",
       "      <td>931</td>\n",
       "      <td>21</td>\n",
       "      <td>69</td>\n",
       "      <td>184</td>\n",
       "      <td>22</td>\n",
       "      <td>2357</td>\n",
       "      <td>65</td>\n",
       "      <td>23</td>\n",
       "      <td>1</td>\n",
       "      <td>62</td>\n",
       "      <td>21407</td>\n",
       "      <td>3945</td>\n",
       "      <td>190</td>\n",
       "      <td>4649</td>\n",
       "      <td>1244</td>\n",
       "      <td>10980</td>\n",
       "      <td>4649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>3616</td>\n",
       "      <td>1010</td>\n",
       "      <td>15</td>\n",
       "      <td>13</td>\n",
       "      <td>160</td>\n",
       "      <td>70</td>\n",
       "      <td>380</td>\n",
       "      <td>4</td>\n",
       "      <td>77</td>\n",
       "      <td>8</td>\n",
       "      <td>9</td>\n",
       "      <td>276</td>\n",
       "      <td>138</td>\n",
       "      <td>45</td>\n",
       "      <td>75</td>\n",
       "      <td>19</td>\n",
       "      <td>362</td>\n",
       "      <td>75</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   RESOURCE  MGR_ID  ROLE_ROLLUP_1  ROLE_ROLLUP_2  ROLE_DEPTNAME  ROLE_TITLE  \\\n",
       "0      3050    4440             21             65            319           4   \n",
       "1       644     162             21             69            310          34   \n",
       "2      2706    1679             50             58             14           0   \n",
       "3      2615     931             21             69            184          22   \n",
       "4      3616    1010             15             13            160          70   \n",
       "\n",
       "   ROLE_FAMILY_DESC  ROLE_FAMILY  ROLE_CODE  RESOURCE_counts  MGR_ID_counts  \\\n",
       "0                 7           65          4                3             55   \n",
       "1                62           67         38               30             10   \n",
       "2              2590            3          0                2              3   \n",
       "3              2357           65         23                1             62   \n",
       "4               380            4         77                8              9   \n",
       "\n",
       "   ROLE_ROLLUP_1_counts  ROLE_ROLLUP_2_counts  ROLE_DEPTNAME_counts  \\\n",
       "0                 21407                  4424                    72   \n",
       "1                 21407                  3945                   159   \n",
       "2                   184                   184                   546   \n",
       "3                 21407                  3945                   190   \n",
       "4                   276                   138                    45   \n",
       "\n",
       "   ROLE_TITLE_counts  ROLE_FAMILY_DESC_counts  ROLE_FAMILY_counts  \\\n",
       "0               3583                     6896               10980   \n",
       "1                 81                       12                1287   \n",
       "2               1256                       33                2636   \n",
       "3               4649                     1244               10980   \n",
       "4                 75                       19                 362   \n",
       "\n",
       "   ROLE_CODE_counts  \n",
       "0              3583  \n",
       "1                81  \n",
       "2              1256  \n",
       "3              4649  \n",
       "4                75  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "_uuid": "45e5ed5a55c304dc5d87d2b55791a06935ee2229"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>RESOURCE</th>\n",
       "      <th>MGR_ID</th>\n",
       "      <th>ROLE_ROLLUP_1</th>\n",
       "      <th>ROLE_ROLLUP_2</th>\n",
       "      <th>ROLE_DEPTNAME</th>\n",
       "      <th>ROLE_TITLE</th>\n",
       "      <th>ROLE_FAMILY_DESC</th>\n",
       "      <th>ROLE_FAMILY</th>\n",
       "      <th>ROLE_CODE</th>\n",
       "      <th>RESOURCE_counts</th>\n",
       "      <th>MGR_ID_counts</th>\n",
       "      <th>ROLE_ROLLUP_1_counts</th>\n",
       "      <th>ROLE_ROLLUP_2_counts</th>\n",
       "      <th>ROLE_DEPTNAME_counts</th>\n",
       "      <th>ROLE_TITLE_counts</th>\n",
       "      <th>ROLE_FAMILY_DESC_counts</th>\n",
       "      <th>ROLE_FAMILY_counts</th>\n",
       "      <th>ROLE_CODE_counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>5481</td>\n",
       "      <td>4250</td>\n",
       "      <td>33</td>\n",
       "      <td>38</td>\n",
       "      <td>13</td>\n",
       "      <td>0</td>\n",
       "      <td>34</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>13</td>\n",
       "      <td>5.0</td>\n",
       "      <td>221.0</td>\n",
       "      <td>221.0</td>\n",
       "      <td>1135.0</td>\n",
       "      <td>1256.0</td>\n",
       "      <td>114.0</td>\n",
       "      <td>2636.0</td>\n",
       "      <td>1256.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>3283</td>\n",
       "      <td>655</td>\n",
       "      <td>21</td>\n",
       "      <td>67</td>\n",
       "      <td>70</td>\n",
       "      <td>55</td>\n",
       "      <td>267</td>\n",
       "      <td>13</td>\n",
       "      <td>59</td>\n",
       "      <td>2</td>\n",
       "      <td>17.0</td>\n",
       "      <td>21407.0</td>\n",
       "      <td>2641.0</td>\n",
       "      <td>267.0</td>\n",
       "      <td>121.0</td>\n",
       "      <td>25.0</td>\n",
       "      <td>1294.0</td>\n",
       "      <td>121.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>4992</td>\n",
       "      <td>298</td>\n",
       "      <td>21</td>\n",
       "      <td>65</td>\n",
       "      <td>155</td>\n",
       "      <td>13</td>\n",
       "      <td>2813</td>\n",
       "      <td>62</td>\n",
       "      <td>13</td>\n",
       "      <td>10</td>\n",
       "      <td>52.0</td>\n",
       "      <td>21407.0</td>\n",
       "      <td>4424.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>154.0</td>\n",
       "      <td>224.0</td>\n",
       "      <td>64.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>3746</td>\n",
       "      <td>2297</td>\n",
       "      <td>21</td>\n",
       "      <td>59</td>\n",
       "      <td>53</td>\n",
       "      <td>124</td>\n",
       "      <td>959</td>\n",
       "      <td>28</td>\n",
       "      <td>133</td>\n",
       "      <td>2</td>\n",
       "      <td>45.0</td>\n",
       "      <td>21407.0</td>\n",
       "      <td>2547.0</td>\n",
       "      <td>532.0</td>\n",
       "      <td>289.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>465.0</td>\n",
       "      <td>289.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>3507</td>\n",
       "      <td>3295</td>\n",
       "      <td>21</td>\n",
       "      <td>69</td>\n",
       "      <td>161</td>\n",
       "      <td>27</td>\n",
       "      <td>2799</td>\n",
       "      <td>14</td>\n",
       "      <td>28</td>\n",
       "      <td>204</td>\n",
       "      <td>19.0</td>\n",
       "      <td>21407.0</td>\n",
       "      <td>3945.0</td>\n",
       "      <td>543.0</td>\n",
       "      <td>376.0</td>\n",
       "      <td>222.0</td>\n",
       "      <td>2690.0</td>\n",
       "      <td>376.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id  RESOURCE  MGR_ID  ROLE_ROLLUP_1  ROLE_ROLLUP_2  ROLE_DEPTNAME  \\\n",
       "0   1      5481    4250             33             38             13   \n",
       "1   2      3283     655             21             67             70   \n",
       "2   3      4992     298             21             65            155   \n",
       "3   4      3746    2297             21             59             53   \n",
       "4   5      3507    3295             21             69            161   \n",
       "\n",
       "   ROLE_TITLE  ROLE_FAMILY_DESC  ROLE_FAMILY  ROLE_CODE  RESOURCE_counts  \\\n",
       "0           0                34            3          0               13   \n",
       "1          55               267           13         59                2   \n",
       "2          13              2813           62         13               10   \n",
       "3         124               959           28        133                2   \n",
       "4          27              2799           14         28              204   \n",
       "\n",
       "   MGR_ID_counts  ROLE_ROLLUP_1_counts  ROLE_ROLLUP_2_counts  \\\n",
       "0            5.0                 221.0                 221.0   \n",
       "1           17.0               21407.0                2641.0   \n",
       "2           52.0               21407.0                4424.0   \n",
       "3           45.0               21407.0                2547.0   \n",
       "4           19.0               21407.0                3945.0   \n",
       "\n",
       "   ROLE_DEPTNAME_counts  ROLE_TITLE_counts  ROLE_FAMILY_DESC_counts  \\\n",
       "0                1135.0             1256.0                    114.0   \n",
       "1                 267.0              121.0                     25.0   \n",
       "2                  34.0               64.0                    154.0   \n",
       "3                 532.0              289.0                     30.0   \n",
       "4                 543.0              376.0                    222.0   \n",
       "\n",
       "   ROLE_FAMILY_counts  ROLE_CODE_counts  \n",
       "0              2636.0            1256.0  \n",
       "1              1294.0             121.0  \n",
       "2               224.0              64.0  \n",
       "3               465.0             289.0  \n",
       "4              2690.0             376.0  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xt.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RESOURCE            7518\n",
       "MGR_ID              4243\n",
       "ROLE_ROLLUP_1        128\n",
       "ROLE_ROLLUP_2        177\n",
       "ROLE_DEPTNAME        449\n",
       "ROLE_TITLE           343\n",
       "ROLE_FAMILY_DESC    2358\n",
       "ROLE_FAMILY           67\n",
       "ROLE_CODE            343\n",
       "dtype: int64"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Counting unique values of categorical variables\n",
    "X[categorical_variables].nunique()"
   ]
  },
  {
   "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>RESOURCE_counts</th>\n",
       "      <th>MGR_ID_counts</th>\n",
       "      <th>ROLE_ROLLUP_1_counts</th>\n",
       "      <th>ROLE_ROLLUP_2_counts</th>\n",
       "      <th>ROLE_DEPTNAME_counts</th>\n",
       "      <th>ROLE_TITLE_counts</th>\n",
       "      <th>ROLE_FAMILY_DESC_counts</th>\n",
       "      <th>ROLE_FAMILY_counts</th>\n",
       "      <th>ROLE_CODE_counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>count</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "      <td>32769.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>mean</td>\n",
       "      <td>83.664408</td>\n",
       "      <td>20.219995</td>\n",
       "      <td>14078.222558</td>\n",
       "      <td>1898.867741</td>\n",
       "      <td>260.167384</td>\n",
       "      <td>1394.336629</td>\n",
       "      <td>1597.602643</td>\n",
       "      <td>4597.008178</td>\n",
       "      <td>1394.336629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>std</td>\n",
       "      <td>164.604062</td>\n",
       "      <td>18.611747</td>\n",
       "      <td>10060.598792</td>\n",
       "      <td>1591.903681</td>\n",
       "      <td>253.684148</td>\n",
       "      <td>1692.657706</td>\n",
       "      <td>2747.932995</td>\n",
       "      <td>4598.402211</td>\n",
       "      <td>1692.657706</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>min</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25%</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>7.000000</td>\n",
       "      <td>334.000000</td>\n",
       "      <td>257.000000</td>\n",
       "      <td>83.000000</td>\n",
       "      <td>164.000000</td>\n",
       "      <td>16.000000</td>\n",
       "      <td>783.000000</td>\n",
       "      <td>164.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50%</td>\n",
       "      <td>13.000000</td>\n",
       "      <td>15.000000</td>\n",
       "      <td>21407.000000</td>\n",
       "      <td>1665.000000</td>\n",
       "      <td>162.000000</td>\n",
       "      <td>467.000000</td>\n",
       "      <td>128.000000</td>\n",
       "      <td>2400.000000</td>\n",
       "      <td>467.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75%</td>\n",
       "      <td>72.000000</td>\n",
       "      <td>27.000000</td>\n",
       "      <td>21407.000000</td>\n",
       "      <td>3945.000000</td>\n",
       "      <td>335.000000</td>\n",
       "      <td>3583.000000</td>\n",
       "      <td>670.000000</td>\n",
       "      <td>10980.000000</td>\n",
       "      <td>3583.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>max</td>\n",
       "      <td>839.000000</td>\n",
       "      <td>152.000000</td>\n",
       "      <td>21407.000000</td>\n",
       "      <td>4424.000000</td>\n",
       "      <td>1135.000000</td>\n",
       "      <td>4649.000000</td>\n",
       "      <td>6896.000000</td>\n",
       "      <td>10980.000000</td>\n",
       "      <td>4649.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       RESOURCE_counts  MGR_ID_counts  ROLE_ROLLUP_1_counts  \\\n",
       "count     32769.000000   32769.000000          32769.000000   \n",
       "mean         83.664408      20.219995          14078.222558   \n",
       "std         164.604062      18.611747          10060.598792   \n",
       "min           1.000000       1.000000              1.000000   \n",
       "25%           3.000000       7.000000            334.000000   \n",
       "50%          13.000000      15.000000          21407.000000   \n",
       "75%          72.000000      27.000000          21407.000000   \n",
       "max         839.000000     152.000000          21407.000000   \n",
       "\n",
       "       ROLE_ROLLUP_2_counts  ROLE_DEPTNAME_counts  ROLE_TITLE_counts  \\\n",
       "count          32769.000000          32769.000000       32769.000000   \n",
       "mean            1898.867741            260.167384        1394.336629   \n",
       "std             1591.903681            253.684148        1692.657706   \n",
       "min                1.000000              1.000000           1.000000   \n",
       "25%              257.000000             83.000000         164.000000   \n",
       "50%             1665.000000            162.000000         467.000000   \n",
       "75%             3945.000000            335.000000        3583.000000   \n",
       "max             4424.000000           1135.000000        4649.000000   \n",
       "\n",
       "       ROLE_FAMILY_DESC_counts  ROLE_FAMILY_counts  ROLE_CODE_counts  \n",
       "count             32769.000000        32769.000000      32769.000000  \n",
       "mean               1597.602643         4597.008178       1394.336629  \n",
       "std                2747.932995         4598.402211       1692.657706  \n",
       "min                   1.000000            2.000000          1.000000  \n",
       "25%                  16.000000          783.000000        164.000000  \n",
       "50%                 128.000000         2400.000000        467.000000  \n",
       "75%                 670.000000        10980.000000       3583.000000  \n",
       "max                6896.000000        10980.000000       4649.000000  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Describing numeric variables\n",
    "X[numeric_variables].describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using CatBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_uuid": "7bd78f7c48f077ace3ba8449323a779865561c36"
   },
   "outputs": [],
   "source": [
    "# Initializing a CatBoostClassifier with best parameters\n",
    "best_params = {'bagging_temperature': 0.6,\n",
    "               'border_count': 200,\n",
    "               'depth': 8,\n",
    "               'iterations': 350,\n",
    "               'l2_leaf_reg': 30,\n",
    "               'learning_rate': 0.30,\n",
    "               'random_strength': 0.01,\n",
    "               'scale_pos_weight': 0.48}\n",
    "\n",
    "catb = CatBoostClassifier(**best_params,\n",
    "                          loss_function='Logloss',\n",
    "                          eval_metric = 'AUC',\n",
    "                          nan_mode='Min',\n",
    "                          thread_count=2,\n",
    "                          verbose = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "_uuid": "a1104a58a48933a659d0dc0cc8c804bf4f091e4e"
   },
   "outputs": [],
   "source": [
    "# Setting a 5-fold stratified cross-validation (note: shuffle=True)\n",
    "SEED = 42\n",
    "FOLDS = 5\n",
    "\n",
    "skf = StratifiedKFold(n_splits=FOLDS, shuffle=True, random_state=SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0:\ttest: 0.6331238\tbest: 0.6331238 (0)\ttotal: 168ms\tremaining: 58.8s\n",
      "100:\ttest: 0.8871443\tbest: 0.8871443 (100)\ttotal: 10.9s\tremaining: 26.9s\n",
      "200:\ttest: 0.8862012\tbest: 0.8891918 (158)\ttotal: 22s\tremaining: 16.3s\n",
      "Stopped by overfitting detector  (50 iterations wait)\n",
      "\n",
      "bestTest = 0.889191775\n",
      "bestIteration = 158\n",
      "\n",
      "Shrink model to first 159 iterations.\n",
      "0:\ttest: 0.6448479\tbest: 0.6448479 (0)\ttotal: 111ms\tremaining: 38.9s\n",
      "100:\ttest: 0.8877733\tbest: 0.8919587 (73)\ttotal: 10.5s\tremaining: 25.8s\n",
      "Stopped by overfitting detector  (50 iterations wait)\n",
      "\n",
      "bestTest = 0.8919586618\n",
      "bestIteration = 73\n",
      "\n",
      "Shrink model to first 74 iterations.\n",
      "0:\ttest: 0.6284797\tbest: 0.6284797 (0)\ttotal: 111ms\tremaining: 38.7s\n",
      "100:\ttest: 0.8899182\tbest: 0.8902680 (83)\ttotal: 11s\tremaining: 27.2s\n",
      "Stopped by overfitting detector  (50 iterations wait)\n",
      "\n",
      "bestTest = 0.8902679805\n",
      "bestIteration = 83\n",
      "\n",
      "Shrink model to first 84 iterations.\n",
      "0:\ttest: 0.6505088\tbest: 0.6505088 (0)\ttotal: 108ms\tremaining: 37.8s\n",
      "100:\ttest: 0.8944298\tbest: 0.8956850 (83)\ttotal: 10.4s\tremaining: 25.7s\n",
      "Stopped by overfitting detector  (50 iterations wait)\n",
      "\n",
      "bestTest = 0.8956849859\n",
      "bestIteration = 83\n",
      "\n",
      "Shrink model to first 84 iterations.\n",
      "0:\ttest: 0.6475395\tbest: 0.6475395 (0)\ttotal: 114ms\tremaining: 39.7s\n",
      "Stopped by overfitting detector  (50 iterations wait)\n",
      "\n",
      "bestTest = 0.8772157563\n",
      "bestIteration = 13\n",
      "\n",
      "Shrink model to first 14 iterations.\n"
     ]
    }
   ],
   "source": [
    "# CV interations\n",
    "\n",
    "roc_auc = list()\n",
    "average_precision = list()\n",
    "oof = np.zeros(len(X))\n",
    "best_iteration = list()\n",
    "\n",
    "for train_idx, test_idx in skf.split(X, y):\n",
    "    X_train, y_train = X.iloc[train_idx, :], y[train_idx]\n",
    "    X_test, y_test = X.iloc[test_idx, :], y[test_idx]\n",
    "    \n",
    "    train = Pool(data=X_train, \n",
    "             label=y_train,            \n",
    "             feature_names=list(X_train.columns),\n",
    "             cat_features=categorical_variables)\n",
    "\n",
    "    test = Pool(data=X_test, \n",
    "                label=y_test,\n",
    "                feature_names=list(X_test.columns),\n",
    "                cat_features=categorical_variables)\n",
    "\n",
    "    catb.fit(train,\n",
    "             verbose_eval=100, \n",
    "             early_stopping_rounds=50,\n",
    "             eval_set=test,\n",
    "             use_best_model=True,\n",
    "             #task_type = \"GPU\",\n",
    "             plot=False)\n",
    "    \n",
    "    best_iteration.append(catb.best_iteration_)\n",
    "    preds = catb.predict_proba(X_test)\n",
    "    \n",
    "    oof[test_idx] = preds[:,1]\n",
    "    \n",
    "    roc_auc.append(roc_auc_score(y_true=y_test, y_score=preds[:,1]))\n",
    "    average_precision.append(average_precision_score(y_true=y_test, y_score=preds[:,1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average cv roc auc score 0.889 ± 0.006\n",
      "Average cv roc average precision 0.990 ± 0.001\n",
      "Roc auc score OOF 0.889\n",
      "Average precision OOF 0.989\n"
     ]
    }
   ],
   "source": [
    "print(\"Average cv roc auc score %0.3f ± %0.3f\" % (np.mean(roc_auc), np.std(roc_auc)))\n",
    "print(\"Average cv roc average precision %0.3f ± %0.3f\" % (np.mean(average_precision), np.std(average_precision)))\n",
    "\n",
    "print(\"Roc auc score OOF %0.3f\" % roc_auc_score(y_true=y, y_score=oof))\n",
    "print(\"Average precision OOF %0.3f\" % average_precision_score(y_true=y, y_score=oof))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0:\ttotal: 46.5ms\tremaining: 4.92s\n",
      "100:\ttotal: 3.64s\tremaining: 216ms\n",
      "106:\ttotal: 3.85s\tremaining: 0us\n"
     ]
    }
   ],
   "source": [
    "# Using catboost on all the data for predictions\n",
    "\n",
    "best_params = {'bagging_temperature': 0.6,\n",
    "               'border_count': 200,\n",
    "               'depth': 8,\n",
    "               'iterations': int(np.median(best_iteration) * 1.3),\n",
    "               'l2_leaf_reg': 30,\n",
    "               'learning_rate': 0.30,\n",
    "               'random_strength': 0.01,\n",
    "               'scale_pos_weight': 0.48}\n",
    "\n",
    "catb = CatBoostClassifier(**best_params,\n",
    "                          loss_function='Logloss',\n",
    "                          eval_metric = 'AUC',\n",
    "                          nan_mode='Min',\n",
    "                          thread_count=2,\n",
    "                          verbose = False)\n",
    "\n",
    "train = Pool(data=X, \n",
    "             label=y,            \n",
    "             feature_names=list(X_train.columns),\n",
    "             cat_features=categorical_variables)\n",
    "\n",
    "catb.fit(train,\n",
    "         verbose_eval=100,\n",
    "         #task_type = \"GPU\",\n",
    "         plot=False)\n",
    "\n",
    "submission = pd.DataFrame(Xt.id)\n",
    "Xt_pool = Pool(data=Xt[list(X_train.columns)],\n",
    "               feature_names=list(X_train.columns),\n",
    "               cat_features=categorical_variables)\n",
    "submission['Action'] = catb.predict_proba(Xt_pool)[:,1]\n",
    "submission.to_csv(\"catboost_submission.csv\", index=False)\n",
    "\n",
    "cat_boost_submission = submission.copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Using deep learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras import backend as K\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.optimizers import Adam, Nadam\n",
    "from tensorflow.keras.layers import Input, Embedding, Reshape, GlobalAveragePooling1D\n",
    "from tensorflow.keras.layers import Flatten, concatenate, Concatenate, Lambda, Dropout, SpatialDropout1D\n",
    "from tensorflow.keras.layers import Reshape, MaxPooling1D,BatchNormalization, AveragePooling1D, Conv1D\n",
    "from tensorflow.keras.layers import Activation, LeakyReLU\n",
    "from tensorflow.keras.optimizers import SGD, Adam, Nadam\n",
    "from tensorflow.keras.models import Model, load_model\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau\n",
    "from tensorflow.keras.regularizers import l2, l1_l2\n",
    "from keras.losses import binary_crossentropy\n",
    "\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.metrics import average_precision_score\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Registering custom activations suitable for tabular problems\n",
    "\n",
    "from tensorflow.keras.utils import get_custom_objects\n",
    "from tensorflow.keras.layers import Activation, LeakyReLU\n",
    "from deep_learning_for_tabular_data.tabular import gelu, Mish, mish\n",
    "\n",
    "# Add gelu so we can use it as a string\n",
    "get_custom_objects().update({'gelu': Activation(gelu)})\n",
    "\n",
    "# Add mish so we can use it as a string\n",
    "get_custom_objects().update({'mish': Mish(mish)})\n",
    "\n",
    "# Add leaky-relu so we can use it as a string\n",
    "get_custom_objects().update({'leaky-relu': Activation(LeakyReLU(alpha=0.2))})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Parametric architecture\n",
    "\n",
    "def tabular_dnn(numeric_variables, categorical_variables, categorical_counts,\n",
    "                feature_selection_dropout=0.2, categorical_dropout=0.1,\n",
    "                first_dense = 256, second_dense = 256, dense_dropout = 0.2, \n",
    "                activation_type=gelu):\n",
    "    \n",
    "    numerical_inputs = Input(shape=(len(numeric_variables),))\n",
    "    numerical_normalization = BatchNormalization()(numerical_inputs)\n",
    "    numerical_feature_selection = Dropout(feature_selection_dropout)(numerical_normalization)\n",
    "\n",
    "    categorical_inputs = []\n",
    "    categorical_embeddings = []\n",
    "    for category in  categorical_variables:\n",
    "        categorical_inputs.append(Input(shape=[1], name=category))\n",
    "        category_counts = categorical_counts[category]\n",
    "        categorical_embeddings.append(\n",
    "            Embedding(category_counts+1, \n",
    "                      int(np.log1p(category_counts)+1), \n",
    "                      name = category + \"_embed\")(categorical_inputs[-1]))\n",
    "\n",
    "    categorical_logits = Concatenate(name = \"categorical_conc\")([Flatten()(SpatialDropout1D(categorical_dropout)(cat_emb)) \n",
    "                                                                 for cat_emb in categorical_embeddings])\n",
    "\n",
    "    x = concatenate([numerical_feature_selection, categorical_logits])\n",
    "    x = Dense(first_dense, activation=activation_type)(x)\n",
    "    x = Dropout(dense_dropout)(x)  \n",
    "    x = Dense(second_dense, activation=activation_type)(x)\n",
    "    x = Dropout(dense_dropout)(x)\n",
    "    output = Dense(1, activation=\"sigmoid\")(x)\n",
    "    model = Model([numerical_inputs] + categorical_inputs, output)\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Useful functions\n",
    "\n",
    "from tensorflow.keras.metrics import AUC\n",
    "\n",
    "def mAP(y_true, y_pred):\n",
    "    return tf.py_function(average_precision_score, (y_true, y_pred), tf.double)\n",
    "\n",
    "def compile_model(model, loss, metrics, optimizer):\n",
    "    model.compile(loss=loss, metrics=metrics, optimizer=optimizer)\n",
    "    return model\n",
    "\n",
    "def plot_keras_history(history, measures):\n",
    "    \"\"\"\n",
    "    history: Keras training history\n",
    "    measures = list of names of measures\n",
    "    \"\"\"\n",
    "    rows = len(measures) // 2 + len(measures) % 2\n",
    "    fig, panels = plt.subplots(rows, 2, figsize=(15, 5))\n",
    "    plt.subplots_adjust(top = 0.99, bottom=0.01, hspace=0.4, wspace=0.2)\n",
    "    try:\n",
    "        panels = [item for sublist in panels for item in sublist]\n",
    "    except:\n",
    "        pass\n",
    "    for k, measure in enumerate(measures):\n",
    "        panel = panels[k]\n",
    "        panel.set_title(measure + ' history')\n",
    "        panel.plot(history.epoch, history.history[measure], label=\"Train \"+measure)\n",
    "        panel.plot(history.epoch, history.history[\"val_\"+measure], label=\"Validation \"+measure)\n",
    "        panel.set(xlabel='epochs', ylabel=measure)\n",
    "        panel.legend()\n",
    "        \n",
    "    plt.show(fig)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Global training settings\n",
    "\n",
    "SEED = 42\n",
    "FOLDS = 5\n",
    "BATCH_SIZE = 512"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Defining callbacks\n",
    "\n",
    "measure_to_monitor = 'val_auc' \n",
    "modality = 'max'\n",
    "\n",
    "early_stopping = EarlyStopping(monitor=measure_to_monitor, \n",
    "                               mode=modality, \n",
    "                               patience=3, \n",
    "                               verbose=0)\n",
    "\n",
    "model_checkpoint = ModelCheckpoint('best.model', \n",
    "                                   monitor=measure_to_monitor, \n",
    "                                   mode=modality, \n",
    "                                   save_best_only=True, \n",
    "                                   verbose=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input array sizes: [9, 6720, 4069, 130, 178, 444, 339, 2221, 69, 339]\n",
      "Categorical levels: {'RESOURCE': 6720, 'MGR_ID': 4069, 'ROLE_ROLLUP_1': 130, 'ROLE_ROLLUP_2': 178, 'ROLE_DEPTNAME': 444, 'ROLE_TITLE': 339, 'ROLE_FAMILY_DESC': 2221, 'ROLE_FAMILY': 69, 'ROLE_CODE': 339}\n",
      "\n",
      "Epoch 1/30\n",
      "52/52 [==============================] - 8s 158ms/step - loss: 0.6037 - auc: 0.4941 - mAP: 0.9423 - val_loss: 0.5247 - val_auc: 0.4979 - val_mAP: 0.9419\n",
      "Epoch 2/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.4350 - auc: 0.5082 - mAP: 0.9435 - val_loss: 0.3638 - val_auc: 0.5151 - val_mAP: 0.9454\n",
      "Epoch 3/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.3025 - auc: 0.5225 - mAP: 0.9463 - val_loss: 0.2604 - val_auc: 0.5442 - val_mAP: 0.9507\n",
      "Epoch 4/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.2392 - auc: 0.5702 - mAP: 0.9528 - val_loss: 0.2256 - val_auc: 0.5873 - val_mAP: 0.9571\n",
      "Epoch 5/30\n",
      "52/52 [==============================] - 4s 67ms/step - loss: 0.2225 - auc: 0.6155 - mAP: 0.9585 - val_loss: 0.2150 - val_auc: 0.6437 - val_mAP: 0.9638\n",
      "Epoch 6/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2148 - auc: 0.6666 - mAP: 0.9645 - val_loss: 0.2089 - val_auc: 0.6869 - val_mAP: 0.9693\n",
      "Epoch 7/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.2047 - auc: 0.7099 - mAP: 0.9709 - val_loss: 0.2032 - val_auc: 0.7300 - val_mAP: 0.9736\n",
      "Epoch 8/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1990 - auc: 0.7562 - mAP: 0.9754 - val_loss: 0.1975 - val_auc: 0.7651 - val_mAP: 0.9774\n",
      "Epoch 9/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1890 - auc: 0.7997 - mAP: 0.9807 - val_loss: 0.1920 - val_auc: 0.7901 - val_mAP: 0.9801\n",
      "Epoch 10/30\n",
      "52/52 [==============================] - 4s 67ms/step - loss: 0.1809 - auc: 0.8356 - mAP: 0.9845 - val_loss: 0.1866 - val_auc: 0.8141 - val_mAP: 0.9824\n",
      "Epoch 11/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1712 - auc: 0.8736 - mAP: 0.9888 - val_loss: 0.1809 - val_auc: 0.8257 - val_mAP: 0.9834\n",
      "Epoch 12/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.1606 - auc: 0.8903 - mAP: 0.9905 - val_loss: 0.1758 - val_auc: 0.8381 - val_mAP: 0.9844\n",
      "Epoch 13/30\n",
      "52/52 [==============================] - 4s 86ms/step - loss: 0.1505 - auc: 0.9129 - mAP: 0.9930 - val_loss: 0.1702 - val_auc: 0.8472 - val_mAP: 0.9854\n",
      "Epoch 14/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1413 - auc: 0.9269 - mAP: 0.9943 - val_loss: 0.1682 - val_auc: 0.8518 - val_mAP: 0.9853\n",
      "Epoch 15/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1309 - auc: 0.9391 - mAP: 0.9956 - val_loss: 0.1654 - val_auc: 0.8537 - val_mAP: 0.9854\n",
      "Epoch 16/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1228 - auc: 0.9491 - mAP: 0.9963 - val_loss: 0.1643 - val_auc: 0.8564 - val_mAP: 0.9856\n",
      "Epoch 17/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1135 - auc: 0.9551 - mAP: 0.9968 - val_loss: 0.1643 - val_auc: 0.8558 - val_mAP: 0.9856\n",
      "Epoch 18/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.1061 - auc: 0.9615 - mAP: 0.9974 - val_loss: 0.1658 - val_auc: 0.8558 - val_mAP: 0.9857\n",
      "Epoch 19/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.1014 - auc: 0.9679 - mAP: 0.9976 - val_loss: 0.1680 - val_auc: 0.8572 - val_mAP: 0.9856\n",
      "Epoch 20/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0943 - auc: 0.9711 - mAP: 0.9981 - val_loss: 0.1697 - val_auc: 0.8556 - val_mAP: 0.9857\n",
      "Epoch 21/30\n",
      "52/52 [==============================] - 4s 79ms/step - loss: 0.0884 - auc: 0.9751 - mAP: 0.9984 - val_loss: 0.1720 - val_auc: 0.8533 - val_mAP: 0.9864\n",
      "Epoch 22/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.0827 - auc: 0.9776 - mAP: 0.9986 - val_loss: 0.1764 - val_auc: 0.8472 - val_mAP: 0.9856\n",
      "\n",
      "FOLD 0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6555/6555 [==============================] - 0s 57us/sample\n",
      "Input array sizes: [9, 6696, 4043, 127, 175, 442, 335, 2217, 69, 335]\n",
      "Categorical levels: {'RESOURCE': 6696, 'MGR_ID': 4043, 'ROLE_ROLLUP_1': 127, 'ROLE_ROLLUP_2': 175, 'ROLE_DEPTNAME': 442, 'ROLE_TITLE': 335, 'ROLE_FAMILY_DESC': 2217, 'ROLE_FAMILY': 69, 'ROLE_CODE': 335}\n",
      "\n",
      "Epoch 1/30\n",
      "52/52 [==============================] - 7s 136ms/step - loss: 0.5888 - auc: 0.5132 - mAP: 0.9401 - val_loss: 0.5068 - val_auc: 0.5047 - val_mAP: 0.9423\n",
      "Epoch 2/30\n",
      "52/52 [==============================] - 3s 64ms/step - loss: 0.4181 - auc: 0.5067 - mAP: 0.9438 - val_loss: 0.3489 - val_auc: 0.5186 - val_mAP: 0.9447\n",
      "Epoch 3/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.2913 - auc: 0.5302 - mAP: 0.9480 - val_loss: 0.2570 - val_auc: 0.5427 - val_mAP: 0.9485\n",
      "Epoch 4/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2395 - auc: 0.5626 - mAP: 0.9520 - val_loss: 0.2282 - val_auc: 0.5809 - val_mAP: 0.9538\n",
      "Epoch 5/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2235 - auc: 0.6109 - mAP: 0.9580 - val_loss: 0.2188 - val_auc: 0.6230 - val_mAP: 0.9594\n",
      "Epoch 6/30\n",
      "52/52 [==============================] - 4s 67ms/step - loss: 0.2132 - auc: 0.6623 - mAP: 0.9648 - val_loss: 0.2131 - val_auc: 0.6623 - val_mAP: 0.9642\n",
      "Epoch 7/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.2051 - auc: 0.7112 - mAP: 0.9704 - val_loss: 0.2078 - val_auc: 0.6940 - val_mAP: 0.9684\n",
      "Epoch 8/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1992 - auc: 0.7533 - mAP: 0.9757 - val_loss: 0.2018 - val_auc: 0.7296 - val_mAP: 0.9728\n",
      "Epoch 9/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1923 - auc: 0.7896 - mAP: 0.9793 - val_loss: 0.1959 - val_auc: 0.7609 - val_mAP: 0.9766\n",
      "Epoch 10/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.1803 - auc: 0.8357 - mAP: 0.9847 - val_loss: 0.1899 - val_auc: 0.7848 - val_mAP: 0.9793\n",
      "Epoch 11/30\n",
      "52/52 [==============================] - 5s 103ms/step - loss: 0.1713 - auc: 0.8670 - mAP: 0.9880 - val_loss: 0.1845 - val_auc: 0.8000 - val_mAP: 0.9811\n",
      "Epoch 12/30\n",
      "52/52 [==============================] - 4s 67ms/step - loss: 0.1627 - auc: 0.8921 - mAP: 0.9907 - val_loss: 0.1788 - val_auc: 0.8155 - val_mAP: 0.9828\n",
      "Epoch 13/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1513 - auc: 0.9142 - mAP: 0.9922 - val_loss: 0.1736 - val_auc: 0.8287 - val_mAP: 0.9842\n",
      "Epoch 14/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1410 - auc: 0.9291 - mAP: 0.9944 - val_loss: 0.1694 - val_auc: 0.8355 - val_mAP: 0.9851\n",
      "Epoch 15/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1308 - auc: 0.9407 - mAP: 0.9954 - val_loss: 0.1656 - val_auc: 0.8445 - val_mAP: 0.9861\n",
      "Epoch 16/30\n",
      "52/52 [==============================] - 3s 64ms/step - loss: 0.1208 - auc: 0.9498 - mAP: 0.9963 - val_loss: 0.1646 - val_auc: 0.8436 - val_mAP: 0.9861\n",
      "Epoch 17/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1147 - auc: 0.9578 - mAP: 0.9969 - val_loss: 0.1625 - val_auc: 0.8504 - val_mAP: 0.9869\n",
      "Epoch 18/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1048 - auc: 0.9640 - mAP: 0.9974 - val_loss: 0.1628 - val_auc: 0.8508 - val_mAP: 0.9871\n",
      "Epoch 19/30\n",
      "52/52 [==============================] - 4s 79ms/step - loss: 0.0967 - auc: 0.9694 - mAP: 0.9980 - val_loss: 0.1626 - val_auc: 0.8523 - val_mAP: 0.9876\n",
      "Epoch 20/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.0913 - auc: 0.9733 - mAP: 0.9983 - val_loss: 0.1660 - val_auc: 0.8506 - val_mAP: 0.9875\n",
      "Epoch 21/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.0880 - auc: 0.9750 - mAP: 0.9982 - val_loss: 0.1686 - val_auc: 0.8483 - val_mAP: 0.9875\n",
      "Epoch 22/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.0817 - auc: 0.9777 - mAP: 0.9986 - val_loss: 0.1725 - val_auc: 0.8407 - val_mAP: 0.9875\n",
      "\n",
      "FOLD 1\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6555/6555 [==============================] - 0s 63us/sample\n",
      "Input array sizes: [9, 6740, 4061, 129, 176, 448, 340, 2226, 69, 340]\n",
      "Categorical levels: {'RESOURCE': 6740, 'MGR_ID': 4061, 'ROLE_ROLLUP_1': 129, 'ROLE_ROLLUP_2': 176, 'ROLE_DEPTNAME': 448, 'ROLE_TITLE': 340, 'ROLE_FAMILY_DESC': 2226, 'ROLE_FAMILY': 69, 'ROLE_CODE': 340}\n",
      "\n",
      "Epoch 1/30\n",
      "52/52 [==============================] - 7s 143ms/step - loss: 0.6083 - auc: 0.4974 - mAP: 0.9395 - val_loss: 0.5243 - val_auc: 0.5020 - val_mAP: 0.9443\n",
      "Epoch 2/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.4392 - auc: 0.5035 - mAP: 0.9426 - val_loss: 0.3664 - val_auc: 0.5120 - val_mAP: 0.9476\n",
      "Epoch 3/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.3073 - auc: 0.5156 - mAP: 0.9454 - val_loss: 0.2629 - val_auc: 0.5373 - val_mAP: 0.9518\n",
      "Epoch 4/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.2445 - auc: 0.5508 - mAP: 0.9506 - val_loss: 0.2267 - val_auc: 0.5878 - val_mAP: 0.9580\n",
      "Epoch 5/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.2245 - auc: 0.5927 - mAP: 0.9569 - val_loss: 0.2156 - val_auc: 0.6353 - val_mAP: 0.9633\n",
      "Epoch 6/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.2171 - auc: 0.6538 - mAP: 0.9630 - val_loss: 0.2096 - val_auc: 0.6796 - val_mAP: 0.9683\n",
      "Epoch 7/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2093 - auc: 0.6911 - mAP: 0.9679 - val_loss: 0.2041 - val_auc: 0.7194 - val_mAP: 0.9730\n",
      "Epoch 8/30\n",
      "52/52 [==============================] - 3s 65ms/step - loss: 0.1994 - auc: 0.7430 - mAP: 0.9744 - val_loss: 0.1989 - val_auc: 0.7513 - val_mAP: 0.9767\n",
      "Epoch 9/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1933 - auc: 0.7880 - mAP: 0.9792 - val_loss: 0.1937 - val_auc: 0.7823 - val_mAP: 0.9801\n",
      "Epoch 10/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1835 - auc: 0.8239 - mAP: 0.9835 - val_loss: 0.1886 - val_auc: 0.8080 - val_mAP: 0.9825\n",
      "Epoch 11/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1756 - auc: 0.8546 - mAP: 0.9864 - val_loss: 0.1832 - val_auc: 0.8248 - val_mAP: 0.9841\n",
      "Epoch 12/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1653 - auc: 0.8808 - mAP: 0.9893 - val_loss: 0.1780 - val_auc: 0.8392 - val_mAP: 0.9852\n",
      "Epoch 13/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1563 - auc: 0.9030 - mAP: 0.9921 - val_loss: 0.1730 - val_auc: 0.8498 - val_mAP: 0.9860\n",
      "Epoch 14/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1463 - auc: 0.9209 - mAP: 0.9933 - val_loss: 0.1688 - val_auc: 0.8550 - val_mAP: 0.9863\n",
      "Epoch 15/30\n",
      "52/52 [==============================] - 5s 106ms/step - loss: 0.1366 - auc: 0.9350 - mAP: 0.9949 - val_loss: 0.1653 - val_auc: 0.8584 - val_mAP: 0.9867\n",
      "Epoch 16/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.1252 - auc: 0.9461 - mAP: 0.9960 - val_loss: 0.1628 - val_auc: 0.8612 - val_mAP: 0.9868\n",
      "Epoch 17/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1170 - auc: 0.9520 - mAP: 0.9966 - val_loss: 0.1614 - val_auc: 0.8612 - val_mAP: 0.9869\n",
      "Epoch 18/30\n",
      "52/52 [==============================] - 4s 80ms/step - loss: 0.1091 - auc: 0.9585 - mAP: 0.9970 - val_loss: 0.1603 - val_auc: 0.8626 - val_mAP: 0.9875\n",
      "Epoch 19/30\n",
      "52/52 [==============================] - 4s 71ms/step - loss: 0.1016 - auc: 0.9642 - mAP: 0.9975 - val_loss: 0.1628 - val_auc: 0.8612 - val_mAP: 0.9869\n",
      "Epoch 20/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0951 - auc: 0.9700 - mAP: 0.9980 - val_loss: 0.1643 - val_auc: 0.8601 - val_mAP: 0.9870\n",
      "Epoch 21/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0895 - auc: 0.9724 - mAP: 0.9982 - val_loss: 0.1665 - val_auc: 0.8576 - val_mAP: 0.9869\n",
      "\n",
      "FOLD 2\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6553/6553 [==============================] - 0s 70us/sample\n",
      "Input array sizes: [9, 6698, 4011, 128, 175, 442, 339, 2201, 69, 339]\n",
      "Categorical levels: {'RESOURCE': 6698, 'MGR_ID': 4011, 'ROLE_ROLLUP_1': 128, 'ROLE_ROLLUP_2': 175, 'ROLE_DEPTNAME': 442, 'ROLE_TITLE': 339, 'ROLE_FAMILY_DESC': 2201, 'ROLE_FAMILY': 69, 'ROLE_CODE': 339}\n",
      "\n",
      "Epoch 1/30\n",
      "52/52 [==============================] - 8s 150ms/step - loss: 0.6022 - auc: 0.4817 - mAP: 0.9405 - val_loss: 0.5201 - val_auc: 0.5044 - val_mAP: 0.9446\n",
      "Epoch 2/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.4364 - auc: 0.4888 - mAP: 0.9416 - val_loss: 0.3626 - val_auc: 0.5388 - val_mAP: 0.9494\n",
      "Epoch 3/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.3064 - auc: 0.5031 - mAP: 0.9432 - val_loss: 0.2597 - val_auc: 0.5813 - val_mAP: 0.9556\n",
      "Epoch 4/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.2425 - auc: 0.5493 - mAP: 0.9520 - val_loss: 0.2239 - val_auc: 0.6278 - val_mAP: 0.9612\n",
      "Epoch 5/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.2239 - auc: 0.6027 - mAP: 0.9574 - val_loss: 0.2137 - val_auc: 0.6705 - val_mAP: 0.9664\n",
      "Epoch 6/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.2108 - auc: 0.6666 - mAP: 0.9662 - val_loss: 0.2074 - val_auc: 0.7096 - val_mAP: 0.9708\n",
      "Epoch 7/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2059 - auc: 0.7120 - mAP: 0.9706 - val_loss: 0.2019 - val_auc: 0.7442 - val_mAP: 0.9743\n",
      "Epoch 8/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1966 - auc: 0.7626 - mAP: 0.9770 - val_loss: 0.1966 - val_auc: 0.7702 - val_mAP: 0.9766\n",
      "Epoch 9/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1883 - auc: 0.8060 - mAP: 0.9816 - val_loss: 0.1915 - val_auc: 0.7868 - val_mAP: 0.9781\n",
      "Epoch 10/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1780 - auc: 0.8461 - mAP: 0.9861 - val_loss: 0.1861 - val_auc: 0.8016 - val_mAP: 0.9800\n",
      "Epoch 11/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1678 - auc: 0.8791 - mAP: 0.9892 - val_loss: 0.1808 - val_auc: 0.8137 - val_mAP: 0.9813\n",
      "Epoch 12/30\n",
      "52/52 [==============================] - 4s 77ms/step - loss: 0.1586 - auc: 0.8995 - mAP: 0.9914 - val_loss: 0.1759 - val_auc: 0.8210 - val_mAP: 0.9823\n",
      "Epoch 13/30\n",
      "52/52 [==============================] - 4s 74ms/step - loss: 0.1477 - auc: 0.9207 - mAP: 0.9934 - val_loss: 0.1719 - val_auc: 0.8266 - val_mAP: 0.9830\n",
      "Epoch 14/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.1366 - auc: 0.9329 - mAP: 0.9948 - val_loss: 0.1682 - val_auc: 0.8366 - val_mAP: 0.9841\n",
      "Epoch 15/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.1260 - auc: 0.9465 - mAP: 0.9963 - val_loss: 0.1662 - val_auc: 0.8385 - val_mAP: 0.9845\n",
      "Epoch 16/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1182 - auc: 0.9534 - mAP: 0.9967 - val_loss: 0.1649 - val_auc: 0.8404 - val_mAP: 0.9851\n",
      "Epoch 17/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1098 - auc: 0.9605 - mAP: 0.9972 - val_loss: 0.1644 - val_auc: 0.8426 - val_mAP: 0.9856\n",
      "Epoch 18/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1022 - auc: 0.9651 - mAP: 0.9976 - val_loss: 0.1654 - val_auc: 0.8417 - val_mAP: 0.9860\n",
      "Epoch 19/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.0954 - auc: 0.9710 - mAP: 0.9981 - val_loss: 0.1671 - val_auc: 0.8432 - val_mAP: 0.9861\n",
      "Epoch 20/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0901 - auc: 0.9725 - mAP: 0.9982 - val_loss: 0.1691 - val_auc: 0.8438 - val_mAP: 0.9864\n",
      "Epoch 21/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.0852 - auc: 0.9762 - mAP: 0.9984 - val_loss: 0.1720 - val_auc: 0.8412 - val_mAP: 0.9863\n",
      "Epoch 22/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0805 - auc: 0.9788 - mAP: 0.9987 - val_loss: 0.1753 - val_auc: 0.8413 - val_mAP: 0.9863\n",
      "Epoch 23/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.0757 - auc: 0.9814 - mAP: 0.9988 - val_loss: 0.1799 - val_auc: 0.8389 - val_mAP: 0.9863\n",
      "\n",
      "FOLD 3\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6553/6553 [==============================] - 0s 74us/sample\n",
      "Input array sizes: [9, 6696, 4036, 126, 174, 445, 335, 2213, 68, 335]\n",
      "Categorical levels: {'RESOURCE': 6696, 'MGR_ID': 4036, 'ROLE_ROLLUP_1': 126, 'ROLE_ROLLUP_2': 174, 'ROLE_DEPTNAME': 445, 'ROLE_TITLE': 335, 'ROLE_FAMILY_DESC': 2213, 'ROLE_FAMILY': 68, 'ROLE_CODE': 335}\n",
      "\n",
      "Epoch 1/30\n",
      "52/52 [==============================] - 9s 177ms/step - loss: 0.6123 - auc: 0.4852 - mAP: 0.9399 - val_loss: 0.5336 - val_auc: 0.4494 - val_mAP: 0.9368\n",
      "Epoch 2/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.4438 - auc: 0.4996 - mAP: 0.9431 - val_loss: 0.3725 - val_auc: 0.4802 - val_mAP: 0.9413\n",
      "Epoch 3/30\n",
      "52/52 [==============================] - 4s 72ms/step - loss: 0.3051 - auc: 0.5215 - mAP: 0.9480 - val_loss: 0.2634 - val_auc: 0.5256 - val_mAP: 0.9469\n",
      "Epoch 4/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.2397 - auc: 0.5698 - mAP: 0.9538 - val_loss: 0.2286 - val_auc: 0.5850 - val_mAP: 0.9535\n",
      "Epoch 5/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.2206 - auc: 0.6223 - mAP: 0.9609 - val_loss: 0.2187 - val_auc: 0.6331 - val_mAP: 0.9590\n",
      "Epoch 6/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.2100 - auc: 0.6676 - mAP: 0.9677 - val_loss: 0.2128 - val_auc: 0.6720 - val_mAP: 0.9637\n",
      "Epoch 7/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.2035 - auc: 0.7191 - mAP: 0.9728 - val_loss: 0.2074 - val_auc: 0.7062 - val_mAP: 0.9678\n",
      "Epoch 8/30\n",
      "52/52 [==============================] - 3s 65ms/step - loss: 0.1955 - auc: 0.7657 - mAP: 0.9782 - val_loss: 0.2022 - val_auc: 0.7348 - val_mAP: 0.9712\n",
      "Epoch 9/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1905 - auc: 0.8017 - mAP: 0.9818 - val_loss: 0.1966 - val_auc: 0.7619 - val_mAP: 0.9744\n",
      "Epoch 10/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1816 - auc: 0.8451 - mAP: 0.9858 - val_loss: 0.1910 - val_auc: 0.7828 - val_mAP: 0.9768\n",
      "Epoch 11/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1697 - auc: 0.8743 - mAP: 0.9893 - val_loss: 0.1856 - val_auc: 0.8007 - val_mAP: 0.9788\n",
      "Epoch 12/30\n",
      "52/52 [==============================] - 3s 67ms/step - loss: 0.1614 - auc: 0.8956 - mAP: 0.9914 - val_loss: 0.1806 - val_auc: 0.8136 - val_mAP: 0.9808\n",
      "Epoch 13/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1509 - auc: 0.9163 - mAP: 0.9933 - val_loss: 0.1757 - val_auc: 0.8231 - val_mAP: 0.9822\n",
      "Epoch 14/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1411 - auc: 0.9307 - mAP: 0.9948 - val_loss: 0.1719 - val_auc: 0.8315 - val_mAP: 0.9834\n",
      "Epoch 15/30\n",
      "52/52 [==============================] - 4s 67ms/step - loss: 0.1313 - auc: 0.9405 - mAP: 0.9957 - val_loss: 0.1691 - val_auc: 0.8375 - val_mAP: 0.9841\n",
      "Epoch 16/30\n",
      "52/52 [==============================] - 4s 69ms/step - loss: 0.1219 - auc: 0.9503 - mAP: 0.9965 - val_loss: 0.1670 - val_auc: 0.8403 - val_mAP: 0.9849\n",
      "Epoch 17/30\n",
      "52/52 [==============================] - 3s 66ms/step - loss: 0.1125 - auc: 0.9590 - mAP: 0.9972 - val_loss: 0.1667 - val_auc: 0.8430 - val_mAP: 0.9852\n",
      "Epoch 18/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.1043 - auc: 0.9644 - mAP: 0.9976 - val_loss: 0.1675 - val_auc: 0.8435 - val_mAP: 0.9854\n",
      "Epoch 19/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0980 - auc: 0.9684 - mAP: 0.9979 - val_loss: 0.1687 - val_auc: 0.8442 - val_mAP: 0.9855\n",
      "Epoch 20/30\n",
      "52/52 [==============================] - 4s 72ms/step - loss: 0.0928 - auc: 0.9724 - mAP: 0.9982 - val_loss: 0.1705 - val_auc: 0.8460 - val_mAP: 0.9857\n",
      "Epoch 21/30\n",
      "52/52 [==============================] - 4s 70ms/step - loss: 0.0871 - auc: 0.9759 - mAP: 0.9984 - val_loss: 0.1745 - val_auc: 0.8358 - val_mAP: 0.9855\n",
      "Epoch 22/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0823 - auc: 0.9774 - mAP: 0.9986 - val_loss: 0.1774 - val_auc: 0.8342 - val_mAP: 0.9855\n",
      "Epoch 23/30\n",
      "52/52 [==============================] - 4s 68ms/step - loss: 0.0779 - auc: 0.9800 - mAP: 0.9988 - val_loss: 0.1811 - val_auc: 0.8336 - val_mAP: 0.9855\n",
      "\n",
      "FOLD 4\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6553/6553 [==============================] - 1s 83us/sample\n"
     ]
    }
   ],
   "source": [
    "from deep_learning_for_tabular_data.tabular import TabularTransformer, DataGenerator\n",
    "\n",
    "# Setting the CV strategy\n",
    "skf = StratifiedKFold(n_splits=FOLDS, \n",
    "                      shuffle=True, \n",
    "                      random_state=SEED)\n",
    "\n",
    "# CV Iteration\n",
    "roc_auc = list()\n",
    "average_precision = list()\n",
    "oof = np.zeros(len(X))\n",
    "best_iteration = list()\n",
    "\n",
    "for fold, (train_idx, test_idx) in enumerate(skf.split(X, y)):\n",
    "    \n",
    "    tb = TabularTransformer(numeric = numeric_variables,\n",
    "                        ordinal = [],\n",
    "                        lowcat  = [],\n",
    "                        highcat = categorical_variables)\n",
    "\n",
    "    tb.fit(X.iloc[train_idx])\n",
    "    sizes = tb.shape(X.iloc[train_idx])\n",
    "    categorical_levels = dict(zip(categorical_variables, sizes[1:]))\n",
    "    print(f\"Input array sizes: {sizes}\")\n",
    "    print(f\"Categorical levels: {categorical_levels}\\n\")\n",
    "    \n",
    "    model = tabular_dnn(numeric_variables, categorical_variables,\n",
    "                        categorical_levels, \n",
    "                        feature_selection_dropout=0.1,\n",
    "                        categorical_dropout=0.1,\n",
    "                        first_dense = 256,\n",
    "                        second_dense = 256,\n",
    "                        dense_dropout = 0.1,\n",
    "                        activation_type=gelu)\n",
    "    \n",
    "    model = compile_model(model, binary_crossentropy, [AUC(name='auc'), mAP], Adam(learning_rate=0.0001))\n",
    "    \n",
    "    train_batch = DataGenerator(X.iloc[train_idx], \n",
    "                                y[train_idx],\n",
    "                                tabular_transformer=tb,\n",
    "                                batch_size=BATCH_SIZE,\n",
    "                                shuffle=True)\n",
    "    \n",
    "    history = model.fit(train_batch,\n",
    "                        validation_data=(tb.transform(X.iloc[test_idx]), y[test_idx]),\n",
    "                        epochs=30,\n",
    "                        callbacks=[model_checkpoint, early_stopping],\n",
    "                        class_weight={0:1.0, 1:(np.sum(y==0) / np.sum(y==1))},\n",
    "                        verbose=1)\n",
    "    \n",
    "    print(\"\\nFOLD %i\" % fold)\n",
    "    plot_keras_history(history, measures=['auc', 'loss'])\n",
    "    \n",
    "    best_iteration.append(np.argmax(history.history['val_auc']) + 1)\n",
    "    preds = model.predict(tb.transform(X.iloc[test_idx]),\n",
    "                          verbose=1,\n",
    "                          batch_size=1024).flatten()\n",
    "\n",
    "    oof[test_idx] = preds\n",
    "\n",
    "    roc_auc.append(roc_auc_score(y_true=y[test_idx], y_score=preds))\n",
    "    average_precision.append(average_precision_score(y_true=y[test_idx], y_score=preds))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average cv roc auc score 0.854 ± 0.005\n",
      "Average cv roc average precision 0.986 ± 0.001\n",
      "Roc auc score OOF 0.853\n",
      "Average precision OOF 0.986\n"
     ]
    }
   ],
   "source": [
    "print(\"Average cv roc auc score %0.3f ± %0.3f\" % (np.mean(roc_auc), np.std(roc_auc)))\n",
    "print(\"Average cv roc average precision %0.3f ± %0.3f\" % (np.mean(average_precision), np.std(average_precision)))\n",
    "\n",
    "print(\"Roc auc score OOF %0.3f\" % roc_auc_score(y_true=y, y_score=oof))\n",
    "print(\"Average precision OOF %0.3f\" % average_precision_score(y_true=y, y_score=oof))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input array sizes: [9, 7520, 4245, 130, 179, 451, 345, 2360, 69, 345]\n",
      "Categorical levels: {'RESOURCE': 7520, 'MGR_ID': 4245, 'ROLE_ROLLUP_1': 130, 'ROLE_ROLLUP_2': 179, 'ROLE_DEPTNAME': 451, 'ROLE_TITLE': 345, 'ROLE_FAMILY_DESC': 2360, 'ROLE_FAMILY': 69, 'ROLE_CODE': 345}\n",
      "\n",
      "Epoch 1/19\n",
      "65/65 [==============================] - 9s 135ms/step - loss: 0.6031 - auc: 0.4688 - mAP: 0.9402\n",
      "Epoch 2/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.4030 - auc: 0.4797 - mAP: 0.9422\n",
      "Epoch 3/19\n",
      "65/65 [==============================] - 4s 61ms/step - loss: 0.2686 - auc: 0.5203 - mAP: 0.9483\n",
      "Epoch 4/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.2466 - auc: 0.5732 - mAP: nan\n",
      "Epoch 5/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.2195 - auc: 0.6186 - mAP: 0.9606\n",
      "Epoch 6/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.2299 - auc: 0.6628 - mAP: nan\n",
      "Epoch 7/19\n",
      "65/65 [==============================] - 4s 61ms/step - loss: 0.2051 - auc: 0.7127 - mAP: 0.9717\n",
      "Epoch 8/19\n",
      "65/65 [==============================] - 4s 65ms/step - loss: 0.1976 - auc: 0.7535 - mAP: 0.9763\n",
      "Epoch 9/19\n",
      "65/65 [==============================] - 4s 63ms/step - loss: 0.1904 - auc: 0.7903 - mAP: 0.9807\n",
      "Epoch 10/19\n",
      "65/65 [==============================] - 4s 63ms/step - loss: 0.1838 - auc: 0.8290 - mAP: 0.9842\n",
      "Epoch 11/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.1798 - auc: 0.8594 - mAP: 0.9877\n",
      "Epoch 12/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.1641 - auc: 0.8869 - mAP: 0.9905\n",
      "Epoch 13/19\n",
      "65/65 [==============================] - 4s 60ms/step - loss: 0.1557 - auc: 0.9044 - mAP: 0.9923\n",
      "Epoch 14/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.1450 - auc: 0.9204 - mAP: 0.9938\n",
      "Epoch 15/19\n",
      "65/65 [==============================] - 4s 61ms/step - loss: 0.1343 - auc: 0.9360 - mAP: 0.9953\n",
      "Epoch 16/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.1240 - auc: 0.9467 - mAP: 0.9961\n",
      "Epoch 17/19\n",
      "65/65 [==============================] - 4s 63ms/step - loss: 0.1156 - auc: 0.9531 - mAP: 0.9966\n",
      "Epoch 18/19\n",
      "65/65 [==============================] - 4s 62ms/step - loss: 0.1084 - auc: 0.9596 - mAP: 0.9973\n",
      "Epoch 19/19\n",
      "65/65 [==============================] - 4s 61ms/step - loss: 0.1007 - auc: 0.9645 - mAP: 0.9977\n"
     ]
    }
   ],
   "source": [
    "# We train on all the examples, using a rule of thumb for the number of iterations\n",
    "\n",
    "tb = TabularTransformer(numeric = numeric_variables,\n",
    "                        ordinal = [],\n",
    "                        lowcat  = [],\n",
    "                        highcat = categorical_variables)\n",
    "\n",
    "tb.fit(X)\n",
    "sizes = tb.shape(X)\n",
    "categorical_levels = dict(zip(categorical_variables, sizes[1:]))\n",
    "print(f\"Input array sizes: {sizes}\")\n",
    "print(f\"Categorical levels: {categorical_levels}\\n\")\n",
    "\n",
    "model = tabular_dnn(numeric_variables, categorical_variables,\n",
    "                        categorical_levels, \n",
    "                        feature_selection_dropout=0.1,\n",
    "                        categorical_dropout=0.1,\n",
    "                        first_dense = 256,\n",
    "                        second_dense = 256,\n",
    "                        dense_dropout = 0.1,\n",
    "                        activation_type=gelu)\n",
    "    \n",
    "model = compile_model(model, binary_crossentropy, [AUC(name='auc'), mAP], Adam(learning_rate=0.0001))    \n",
    "\n",
    "train_batch = DataGenerator(X, y,\n",
    "                            tabular_transformer=tb,\n",
    "                            batch_size=BATCH_SIZE,\n",
    "                            shuffle=True)\n",
    "\n",
    "history = model.fit(train_batch,\n",
    "                    epochs=int(np.median(best_iteration)),\n",
    "                    class_weight={0:1.0, 1:(np.sum(y==0) / np.sum(y==1))},\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "58921/58921 [==============================] - 2s 35us/sample\n"
     ]
    }
   ],
   "source": [
    "# Predicting and submission\n",
    "preds = model.predict(tb.transform(Xt[X.columns]),\n",
    "                      verbose=1,\n",
    "                      batch_size=1024).flatten()\n",
    "\n",
    "submission = pd.DataFrame(Xt.id)\n",
    "submission['Action'] = preds\n",
    "submission.to_csv(\"tabular_dnn_submission.csv\", index=False)\n",
    "\n",
    "tabular_dnn_submission = submission.copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Blending together the GBM and DNN solutions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy.stats import rankdata\n",
    "\n",
    "# We use normalized ranks because probabilities emissions from the two models may differ\n",
    "dnn_rank = rankdata(tabular_dnn_submission.Action, method='dense') / len(Xt)\n",
    "cat_rank = rankdata(cat_boost_submission.Action, method='dense') / len(Xt)\n",
    "\n",
    "submission = pd.DataFrame(Xt.id)\n",
    "submission['Action'] = 0.5 * dnn_rank + 0.5 * cat_rank \n",
    "submission.to_csv(\"blended_submission.csv\", index=False)"
   ]
  },
  {
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}