notebooks/cohort/case_study/decoupled_class/case_3.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Decoupled Classifiers Case Study 3\n",
    "\n",
    "This notebook will follow a similar approach to what was done in the notebook [Decoupled Classifiers Case Study 2](./case_2.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append('../../../../notebooks')\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from lightgbm import LGBMClassifier\n",
    "\n",
    "from raimitigations.utils import split_data\n",
    "import raimitigations.dataprocessing as dp\n",
    "from raimitigations.cohort import DecoupledClass, CohortDefinition, CohortManager, fetch_cohort_results, plot_value_counts_cohort\n",
    "from sklearn.pipeline import Pipeline\n",
    "from download import download_datasets\n",
    "\n",
    "SEED = 100"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load and split the data into train and test sets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>department</th>\n",
       "      <th>region</th>\n",
       "      <th>education</th>\n",
       "      <th>gender</th>\n",
       "      <th>recruitment_channel</th>\n",
       "      <th>no_of_trainings</th>\n",
       "      <th>age</th>\n",
       "      <th>previous_year_rating</th>\n",
       "      <th>length_of_service</th>\n",
       "      <th>KPIs_met &gt;80%</th>\n",
       "      <th>awards_won?</th>\n",
       "      <th>avg_training_score</th>\n",
       "      <th>is_promoted</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Sales &amp; Marketing</td>\n",
       "      <td>region_7</td>\n",
       "      <td>Master's &amp; above</td>\n",
       "      <td>f</td>\n",
       "      <td>sourcing</td>\n",
       "      <td>1</td>\n",
       "      <td>35</td>\n",
       "      <td>5.0</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>49</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Operations</td>\n",
       "      <td>region_22</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "      <td>30</td>\n",
       "      <td>5.0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>60</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Sales &amp; Marketing</td>\n",
       "      <td>region_19</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>sourcing</td>\n",
       "      <td>1</td>\n",
       "      <td>34</td>\n",
       "      <td>3.0</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>50</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Sales &amp; Marketing</td>\n",
       "      <td>region_23</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>other</td>\n",
       "      <td>2</td>\n",
       "      <td>39</td>\n",
       "      <td>1.0</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>50</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Technology</td>\n",
       "      <td>region_26</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "      <td>45</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>73</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54803</th>\n",
       "      <td>Technology</td>\n",
       "      <td>region_14</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>sourcing</td>\n",
       "      <td>1</td>\n",
       "      <td>48</td>\n",
       "      <td>3.0</td>\n",
       "      <td>17</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>78</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54804</th>\n",
       "      <td>Operations</td>\n",
       "      <td>region_27</td>\n",
       "      <td>Master's &amp; above</td>\n",
       "      <td>f</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "      <td>37</td>\n",
       "      <td>2.0</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>56</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54805</th>\n",
       "      <td>Analytics</td>\n",
       "      <td>region_1</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>5.0</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>79</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54806</th>\n",
       "      <td>Sales &amp; Marketing</td>\n",
       "      <td>region_9</td>\n",
       "      <td>NaN</td>\n",
       "      <td>m</td>\n",
       "      <td>sourcing</td>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>45</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>54807</th>\n",
       "      <td>HR</td>\n",
       "      <td>region_22</td>\n",
       "      <td>Bachelor's</td>\n",
       "      <td>m</td>\n",
       "      <td>other</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>1.0</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>49</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>54808 rows × 13 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              department     region         education gender  \\\n",
       "0      Sales & Marketing   region_7  Master's & above      f   \n",
       "1             Operations  region_22        Bachelor's      m   \n",
       "2      Sales & Marketing  region_19        Bachelor's      m   \n",
       "3      Sales & Marketing  region_23        Bachelor's      m   \n",
       "4             Technology  region_26        Bachelor's      m   \n",
       "...                  ...        ...               ...    ...   \n",
       "54803         Technology  region_14        Bachelor's      m   \n",
       "54804         Operations  region_27  Master's & above      f   \n",
       "54805          Analytics   region_1        Bachelor's      m   \n",
       "54806  Sales & Marketing   region_9               NaN      m   \n",
       "54807                 HR  region_22        Bachelor's      m   \n",
       "\n",
       "      recruitment_channel  no_of_trainings  age  previous_year_rating  \\\n",
       "0                sourcing                1   35                   5.0   \n",
       "1                   other                1   30                   5.0   \n",
       "2                sourcing                1   34                   3.0   \n",
       "3                   other                2   39                   1.0   \n",
       "4                   other                1   45                   3.0   \n",
       "...                   ...              ...  ...                   ...   \n",
       "54803            sourcing                1   48                   3.0   \n",
       "54804               other                1   37                   2.0   \n",
       "54805               other                1   27                   5.0   \n",
       "54806            sourcing                1   29                   1.0   \n",
       "54807               other                1   27                   1.0   \n",
       "\n",
       "       length_of_service  KPIs_met >80%  awards_won?  avg_training_score  \\\n",
       "0                      8              1            0                  49   \n",
       "1                      4              0            0                  60   \n",
       "2                      7              0            0                  50   \n",
       "3                     10              0            0                  50   \n",
       "4                      2              0            0                  73   \n",
       "...                  ...            ...          ...                 ...   \n",
       "54803                 17              0            0                  78   \n",
       "54804                  6              0            0                  56   \n",
       "54805                  3              1            0                  79   \n",
       "54806                  2              0            0                  45   \n",
       "54807                  5              0            0                  49   \n",
       "\n",
       "       is_promoted  \n",
       "0                0  \n",
       "1                0  \n",
       "2                0  \n",
       "3                0  \n",
       "4                0  \n",
       "...            ...  \n",
       "54803            0  \n",
       "54804            0  \n",
       "54805            0  \n",
       "54806            0  \n",
       "54807            0  \n",
       "\n",
       "[54808 rows x 13 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_dir = '../../../../datasets/'\n",
    "download_datasets(data_dir)\n",
    "df = pd.read_csv(data_dir + 'hr_promotion/train.csv')\n",
    "df.drop(columns=['employee_id'], inplace=True)\n",
    "label_col = 'is_promoted'\n",
    "\n",
    "X_train, X_test, y_train, y_test = split_data(df, label_col, test_size=0.3, random_state=SEED)\n",
    "\n",
    "df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model():\n",
    "    #model = DecisionTreeClassifier(max_features=\"sqrt\")\n",
    "    model = LGBMClassifier(random_state=SEED)\n",
    "    return model"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline Pipeline\n",
    "\n",
    "This dataset is the same one used in the [Cohort Case Study 3](../case_3.ipynb) notebook, and as depicted there, the ``department`` column is the feature where we can identify some disparity between cohorts. Here, we'll once again consider this feature as a sensitive feature.\n",
    "\n",
    "As our first experiment, let's create a simple pipeline for our baseline:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>cohort</th>\n",
       "      <th>cht_query</th>\n",
       "      <th>roc</th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>threshold</th>\n",
       "      <th>num_pos</th>\n",
       "      <th>%_pos</th>\n",
       "      <th>cht_size</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>all</td>\n",
       "      <td>all</td>\n",
       "      <td>0.910026</td>\n",
       "      <td>0.633798</td>\n",
       "      <td>0.808795</td>\n",
       "      <td>0.658027</td>\n",
       "      <td>0.816092</td>\n",
       "      <td>0.122429</td>\n",
       "      <td>3864</td>\n",
       "      <td>0.234994</td>\n",
       "      <td>16443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>cohort_0</td>\n",
       "      <td>(`department` == \"Analytics\")</td>\n",
       "      <td>0.815071</td>\n",
       "      <td>0.615072</td>\n",
       "      <td>0.737173</td>\n",
       "      <td>0.633686</td>\n",
       "      <td>0.803662</td>\n",
       "      <td>0.144258</td>\n",
       "      <td>357</td>\n",
       "      <td>0.225379</td>\n",
       "      <td>1584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cohort_1</td>\n",
       "      <td>(`department` == \"Finance\")</td>\n",
       "      <td>0.928899</td>\n",
       "      <td>0.642252</td>\n",
       "      <td>0.818997</td>\n",
       "      <td>0.660270</td>\n",
       "      <td>0.791501</td>\n",
       "      <td>0.099002</td>\n",
       "      <td>210</td>\n",
       "      <td>0.278884</td>\n",
       "      <td>753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>cohort_2</td>\n",
       "      <td>(`department` == \"HR\")</td>\n",
       "      <td>0.924777</td>\n",
       "      <td>0.602336</td>\n",
       "      <td>0.845885</td>\n",
       "      <td>0.610867</td>\n",
       "      <td>0.794034</td>\n",
       "      <td>0.096400</td>\n",
       "      <td>179</td>\n",
       "      <td>0.254261</td>\n",
       "      <td>704</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>cohort_3</td>\n",
       "      <td>(`department` == \"Legal\")</td>\n",
       "      <td>0.909836</td>\n",
       "      <td>0.579581</td>\n",
       "      <td>0.763503</td>\n",
       "      <td>0.581716</td>\n",
       "      <td>0.787037</td>\n",
       "      <td>0.107423</td>\n",
       "      <td>78</td>\n",
       "      <td>0.240741</td>\n",
       "      <td>324</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>cohort_4</td>\n",
       "      <td>(`department` == \"Operations\")</td>\n",
       "      <td>0.899921</td>\n",
       "      <td>0.621445</td>\n",
       "      <td>0.797078</td>\n",
       "      <td>0.632185</td>\n",
       "      <td>0.777155</td>\n",
       "      <td>0.107618</td>\n",
       "      <td>953</td>\n",
       "      <td>0.281287</td>\n",
       "      <td>3388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>cohort_5</td>\n",
       "      <td>(`department` == \"Procurement\")</td>\n",
       "      <td>0.914249</td>\n",
       "      <td>0.636048</td>\n",
       "      <td>0.823387</td>\n",
       "      <td>0.654546</td>\n",
       "      <td>0.796365</td>\n",
       "      <td>0.137644</td>\n",
       "      <td>576</td>\n",
       "      <td>0.268406</td>\n",
       "      <td>2146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>cohort_6</td>\n",
       "      <td>(`department` == \"R&amp;D\")</td>\n",
       "      <td>0.721571</td>\n",
       "      <td>0.551094</td>\n",
       "      <td>0.633618</td>\n",
       "      <td>0.546714</td>\n",
       "      <td>0.768166</td>\n",
       "      <td>0.113184</td>\n",
       "      <td>66</td>\n",
       "      <td>0.228374</td>\n",
       "      <td>289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>cohort_7</td>\n",
       "      <td>(`department` == \"Sales &amp; Marketing\")</td>\n",
       "      <td>0.945638</td>\n",
       "      <td>0.658135</td>\n",
       "      <td>0.868133</td>\n",
       "      <td>0.698209</td>\n",
       "      <td>0.860826</td>\n",
       "      <td>0.129820</td>\n",
       "      <td>999</td>\n",
       "      <td>0.193642</td>\n",
       "      <td>5159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>cohort_8</td>\n",
       "      <td>(`department` == \"Technology\")</td>\n",
       "      <td>0.879545</td>\n",
       "      <td>0.629866</td>\n",
       "      <td>0.778378</td>\n",
       "      <td>0.644261</td>\n",
       "      <td>0.780057</td>\n",
       "      <td>0.132835</td>\n",
       "      <td>582</td>\n",
       "      <td>0.277672</td>\n",
       "      <td>2096</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cohort                              cht_query       roc  precision  \\\n",
       "0       all                                    all  0.910026   0.633798   \n",
       "1  cohort_0          (`department` == \"Analytics\")  0.815071   0.615072   \n",
       "2  cohort_1            (`department` == \"Finance\")  0.928899   0.642252   \n",
       "3  cohort_2                 (`department` == \"HR\")  0.924777   0.602336   \n",
       "4  cohort_3              (`department` == \"Legal\")  0.909836   0.579581   \n",
       "5  cohort_4         (`department` == \"Operations\")  0.899921   0.621445   \n",
       "6  cohort_5        (`department` == \"Procurement\")  0.914249   0.636048   \n",
       "7  cohort_6                (`department` == \"R&D\")  0.721571   0.551094   \n",
       "8  cohort_7  (`department` == \"Sales & Marketing\")  0.945638   0.658135   \n",
       "9  cohort_8         (`department` == \"Technology\")  0.879545   0.629866   \n",
       "\n",
       "     recall        f1  accuracy  threshold  num_pos     %_pos  cht_size  \n",
       "0  0.808795  0.658027  0.816092   0.122429     3864  0.234994     16443  \n",
       "1  0.737173  0.633686  0.803662   0.144258      357  0.225379      1584  \n",
       "2  0.818997  0.660270  0.791501   0.099002      210  0.278884       753  \n",
       "3  0.845885  0.610867  0.794034   0.096400      179  0.254261       704  \n",
       "4  0.763503  0.581716  0.787037   0.107423       78  0.240741       324  \n",
       "5  0.797078  0.632185  0.777155   0.107618      953  0.281287      3388  \n",
       "6  0.823387  0.654546  0.796365   0.137644      576  0.268406      2146  \n",
       "7  0.633618  0.546714  0.768166   0.113184       66  0.228374       289  \n",
       "8  0.868133  0.698209  0.860826   0.129820      999  0.193642      5159  \n",
       "9  0.778378  0.644261  0.780057   0.132835      582  0.277672      2096  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pipe = Pipeline([\n",
    "    (\"imputer\", dp.BasicImputer(verbose=False)),\n",
    "    (\"scaler\", dp.DataStandardScaler(verbose=False)),\n",
    "    (\"encoder\", dp.EncoderOHE(verbose=False)),\n",
    "    (\"estimator\", get_model())\n",
    "])\n",
    "\n",
    "pipe.fit(X_train, y_train)\n",
    "pred = pipe.predict_proba(X_test)\n",
    "\n",
    "pred_train = pipe.predict_proba(X_train)\n",
    "_, th_dict = fetch_cohort_results(X_train, y_train, pred_train, cohort_col=[\"department\"], return_th_dict=True)\n",
    "fetch_cohort_results(X_test, y_test, pred, cohort_col=[\"department\"], fixed_th=th_dict)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using the CohortManager\n",
    "\n",
    "Let's now see if we can improve our results over each cohort by building a separate pipeline for each cohort using the ``CohortManager`` class, similar to what we did in the [Cohort Case Study 3](../case_3.ipynb) notebook:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cohort</th>\n",
       "      <th>cht_query</th>\n",
       "      <th>roc</th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>threshold</th>\n",
       "      <th>num_pos</th>\n",
       "      <th>%_pos</th>\n",
       "      <th>cht_size</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>all</td>\n",
       "      <td>all</td>\n",
       "      <td>0.897907</td>\n",
       "      <td>0.666922</td>\n",
       "      <td>0.740960</td>\n",
       "      <td>0.693698</td>\n",
       "      <td>0.882807</td>\n",
       "      <td>0.178975</td>\n",
       "      <td>2123</td>\n",
       "      <td>0.129113</td>\n",
       "      <td>16443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>cohort_0</td>\n",
       "      <td>(`department` == \"Analytics\")</td>\n",
       "      <td>0.801956</td>\n",
       "      <td>0.610089</td>\n",
       "      <td>0.668797</td>\n",
       "      <td>0.628000</td>\n",
       "      <td>0.839015</td>\n",
       "      <td>0.169265</td>\n",
       "      <td>243</td>\n",
       "      <td>0.153409</td>\n",
       "      <td>1584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cohort_1</td>\n",
       "      <td>(`department` == \"Finance\")</td>\n",
       "      <td>0.915831</td>\n",
       "      <td>0.853684</td>\n",
       "      <td>0.744484</td>\n",
       "      <td>0.786259</td>\n",
       "      <td>0.934927</td>\n",
       "      <td>0.500199</td>\n",
       "      <td>50</td>\n",
       "      <td>0.066401</td>\n",
       "      <td>753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>cohort_2</td>\n",
       "      <td>(`department` == \"HR\")</td>\n",
       "      <td>0.894188</td>\n",
       "      <td>0.788518</td>\n",
       "      <td>0.682923</td>\n",
       "      <td>0.721861</td>\n",
       "      <td>0.948864</td>\n",
       "      <td>0.420675</td>\n",
       "      <td>26</td>\n",
       "      <td>0.036932</td>\n",
       "      <td>704</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>cohort_3</td>\n",
       "      <td>(`department` == \"Legal\")</td>\n",
       "      <td>0.911130</td>\n",
       "      <td>0.894654</td>\n",
       "      <td>0.629940</td>\n",
       "      <td>0.687961</td>\n",
       "      <td>0.953704</td>\n",
       "      <td>0.807405</td>\n",
       "      <td>6</td>\n",
       "      <td>0.018519</td>\n",
       "      <td>324</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>cohort_4</td>\n",
       "      <td>(`department` == \"Operations\")</td>\n",
       "      <td>0.894509</td>\n",
       "      <td>0.666407</td>\n",
       "      <td>0.751623</td>\n",
       "      <td>0.695174</td>\n",
       "      <td>0.872491</td>\n",
       "      <td>0.162195</td>\n",
       "      <td>496</td>\n",
       "      <td>0.146399</td>\n",
       "      <td>3388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>cohort_5</td>\n",
       "      <td>(`department` == \"Procurement\")</td>\n",
       "      <td>0.898282</td>\n",
       "      <td>0.674369</td>\n",
       "      <td>0.738740</td>\n",
       "      <td>0.698804</td>\n",
       "      <td>0.881640</td>\n",
       "      <td>0.223448</td>\n",
       "      <td>279</td>\n",
       "      <td>0.130009</td>\n",
       "      <td>2146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>cohort_6</td>\n",
       "      <td>(`department` == \"R&amp;D\")</td>\n",
       "      <td>0.588664</td>\n",
       "      <td>0.463415</td>\n",
       "      <td>0.496269</td>\n",
       "      <td>0.479279</td>\n",
       "      <td>0.920415</td>\n",
       "      <td>0.742253</td>\n",
       "      <td>2</td>\n",
       "      <td>0.006920</td>\n",
       "      <td>289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>cohort_7</td>\n",
       "      <td>(`department` == \"Sales &amp; Marketing\")</td>\n",
       "      <td>0.938696</td>\n",
       "      <td>0.670041</td>\n",
       "      <td>0.787935</td>\n",
       "      <td>0.707151</td>\n",
       "      <td>0.893196</td>\n",
       "      <td>0.165917</td>\n",
       "      <td>674</td>\n",
       "      <td>0.130645</td>\n",
       "      <td>5159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>cohort_8</td>\n",
       "      <td>(`department` == \"Technology\")</td>\n",
       "      <td>0.877133</td>\n",
       "      <td>0.662127</td>\n",
       "      <td>0.727013</td>\n",
       "      <td>0.685259</td>\n",
       "      <td>0.857824</td>\n",
       "      <td>0.201117</td>\n",
       "      <td>325</td>\n",
       "      <td>0.155057</td>\n",
       "      <td>2096</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cohort                              cht_query       roc  precision  \\\n",
       "0       all                                    all  0.897907   0.666922   \n",
       "1  cohort_0          (`department` == \"Analytics\")  0.801956   0.610089   \n",
       "2  cohort_1            (`department` == \"Finance\")  0.915831   0.853684   \n",
       "3  cohort_2                 (`department` == \"HR\")  0.894188   0.788518   \n",
       "4  cohort_3              (`department` == \"Legal\")  0.911130   0.894654   \n",
       "5  cohort_4         (`department` == \"Operations\")  0.894509   0.666407   \n",
       "6  cohort_5        (`department` == \"Procurement\")  0.898282   0.674369   \n",
       "7  cohort_6                (`department` == \"R&D\")  0.588664   0.463415   \n",
       "8  cohort_7  (`department` == \"Sales & Marketing\")  0.938696   0.670041   \n",
       "9  cohort_8         (`department` == \"Technology\")  0.877133   0.662127   \n",
       "\n",
       "     recall        f1  accuracy  threshold  num_pos     %_pos  cht_size  \n",
       "0  0.740960  0.693698  0.882807   0.178975     2123  0.129113     16443  \n",
       "1  0.668797  0.628000  0.839015   0.169265      243  0.153409      1584  \n",
       "2  0.744484  0.786259  0.934927   0.500199       50  0.066401       753  \n",
       "3  0.682923  0.721861  0.948864   0.420675       26  0.036932       704  \n",
       "4  0.629940  0.687961  0.953704   0.807405        6  0.018519       324  \n",
       "5  0.751623  0.695174  0.872491   0.162195      496  0.146399      3388  \n",
       "6  0.738740  0.698804  0.881640   0.223448      279  0.130009      2146  \n",
       "7  0.496269  0.479279  0.920415   0.742253        2  0.006920       289  \n",
       "8  0.787935  0.707151  0.893196   0.165917      674  0.130645      5159  \n",
       "9  0.727013  0.685259  0.857824   0.201117      325  0.155057      2096  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cht_manager = CohortManager(\n",
    "    transform_pipe=[\n",
    "        dp.BasicImputer(verbose=False),\n",
    "        dp.DataStandardScaler(verbose=False),\n",
    "        dp.EncoderOHE(drop=False, unknown_err=False, verbose=False),\n",
    "        get_model()\n",
    "    ],\n",
    "    cohort_col=[\"department\"]\n",
    ")\n",
    "cht_manager.fit(X_train, y_train)\n",
    "pred_cht = cht_manager.predict_proba(X_test)\n",
    "\n",
    "pred_train = cht_manager.predict_proba(X_train)\n",
    "metrics_train, th_dict = fetch_cohort_results(X_train, y_train, pred_train, cohort_col=[\"department\"], return_th_dict=True)\n",
    "fetch_cohort_results(X_test, y_test, pred_cht, cohort_col=[\"department\"], fixed_th=th_dict)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that by just training one model for each cohort based on the ``department`` column we can get an improvement over certain metrics. \n",
    "\n",
    "\n",
    "## DecoupledClassifier - Transfer Learning\n",
    "\n",
    "Let's see if we can improve the *\"department\"* cohorts using the `DecoupledClass`. We'll use transfer learning, but for now we won't use any fairness optimization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>cohort</th>\n",
       "      <th>cht_query</th>\n",
       "      <th>roc</th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>threshold</th>\n",
       "      <th>num_pos</th>\n",
       "      <th>%_pos</th>\n",
       "      <th>cht_size</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>all</td>\n",
       "      <td>all</td>\n",
       "      <td>0.901364</td>\n",
       "      <td>0.885156</td>\n",
       "      <td>0.675796</td>\n",
       "      <td>0.733884</td>\n",
       "      <td>0.939001</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>607</td>\n",
       "      <td>0.036915</td>\n",
       "      <td>16443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>cohort_0</td>\n",
       "      <td>(`department` == \"Analytics\")</td>\n",
       "      <td>0.801923</td>\n",
       "      <td>0.648600</td>\n",
       "      <td>0.611458</td>\n",
       "      <td>0.626030</td>\n",
       "      <td>0.888889</td>\n",
       "      <td>0.285810</td>\n",
       "      <td>108</td>\n",
       "      <td>0.068182</td>\n",
       "      <td>1584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cohort_1</td>\n",
       "      <td>(`department` == \"Finance\")</td>\n",
       "      <td>0.928840</td>\n",
       "      <td>0.883305</td>\n",
       "      <td>0.760029</td>\n",
       "      <td>0.806662</td>\n",
       "      <td>0.941567</td>\n",
       "      <td>0.237477</td>\n",
       "      <td>49</td>\n",
       "      <td>0.065073</td>\n",
       "      <td>753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>cohort_2</td>\n",
       "      <td>(`department` == \"HR\")</td>\n",
       "      <td>0.910229</td>\n",
       "      <td>0.803801</td>\n",
       "      <td>0.649475</td>\n",
       "      <td>0.696304</td>\n",
       "      <td>0.948864</td>\n",
       "      <td>0.237990</td>\n",
       "      <td>20</td>\n",
       "      <td>0.028409</td>\n",
       "      <td>704</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>cohort_3</td>\n",
       "      <td>(`department` == \"Legal\")</td>\n",
       "      <td>0.917688</td>\n",
       "      <td>0.733660</td>\n",
       "      <td>0.722088</td>\n",
       "      <td>0.727695</td>\n",
       "      <td>0.941358</td>\n",
       "      <td>0.227371</td>\n",
       "      <td>18</td>\n",
       "      <td>0.055556</td>\n",
       "      <td>324</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>cohort_4</td>\n",
       "      <td>(`department` == \"Operations\")</td>\n",
       "      <td>0.894497</td>\n",
       "      <td>0.756851</td>\n",
       "      <td>0.721753</td>\n",
       "      <td>0.737574</td>\n",
       "      <td>0.919126</td>\n",
       "      <td>0.270537</td>\n",
       "      <td>262</td>\n",
       "      <td>0.077332</td>\n",
       "      <td>3388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>cohort_5</td>\n",
       "      <td>(`department` == \"Procurement\")</td>\n",
       "      <td>0.898306</td>\n",
       "      <td>0.752563</td>\n",
       "      <td>0.721805</td>\n",
       "      <td>0.735838</td>\n",
       "      <td>0.917987</td>\n",
       "      <td>0.342704</td>\n",
       "      <td>169</td>\n",
       "      <td>0.078751</td>\n",
       "      <td>2146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>cohort_6</td>\n",
       "      <td>(`department` == \"R&amp;D\")</td>\n",
       "      <td>0.699893</td>\n",
       "      <td>0.545614</td>\n",
       "      <td>0.541578</td>\n",
       "      <td>0.543401</td>\n",
       "      <td>0.882353</td>\n",
       "      <td>0.226767</td>\n",
       "      <td>19</td>\n",
       "      <td>0.065744</td>\n",
       "      <td>289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>cohort_7</td>\n",
       "      <td>(`department` == \"Sales &amp; Marketing\")</td>\n",
       "      <td>0.938689</td>\n",
       "      <td>0.777230</td>\n",
       "      <td>0.739770</td>\n",
       "      <td>0.756811</td>\n",
       "      <td>0.939135</td>\n",
       "      <td>0.272814</td>\n",
       "      <td>319</td>\n",
       "      <td>0.061834</td>\n",
       "      <td>5159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>cohort_8</td>\n",
       "      <td>(`department` == \"Technology\")</td>\n",
       "      <td>0.877133</td>\n",
       "      <td>0.683276</td>\n",
       "      <td>0.678080</td>\n",
       "      <td>0.680620</td>\n",
       "      <td>0.882156</td>\n",
       "      <td>0.299429</td>\n",
       "      <td>212</td>\n",
       "      <td>0.101145</td>\n",
       "      <td>2096</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cohort                              cht_query       roc  precision  \\\n",
       "0       all                                    all  0.901364   0.885156   \n",
       "1  cohort_0          (`department` == \"Analytics\")  0.801923   0.648600   \n",
       "2  cohort_1            (`department` == \"Finance\")  0.928840   0.883305   \n",
       "3  cohort_2                 (`department` == \"HR\")  0.910229   0.803801   \n",
       "4  cohort_3              (`department` == \"Legal\")  0.917688   0.733660   \n",
       "5  cohort_4         (`department` == \"Operations\")  0.894497   0.756851   \n",
       "6  cohort_5        (`department` == \"Procurement\")  0.898306   0.752563   \n",
       "7  cohort_6                (`department` == \"R&D\")  0.699893   0.545614   \n",
       "8  cohort_7  (`department` == \"Sales & Marketing\")  0.938689   0.777230   \n",
       "9  cohort_8         (`department` == \"Technology\")  0.877133   0.683276   \n",
       "\n",
       "     recall        f1  accuracy  threshold  num_pos     %_pos  cht_size  \n",
       "0  0.675796  0.733884  0.939001   0.500000      607  0.036915     16443  \n",
       "1  0.611458  0.626030  0.888889   0.285810      108  0.068182      1584  \n",
       "2  0.760029  0.806662  0.941567   0.237477       49  0.065073       753  \n",
       "3  0.649475  0.696304  0.948864   0.237990       20  0.028409       704  \n",
       "4  0.722088  0.727695  0.941358   0.227371       18  0.055556       324  \n",
       "5  0.721753  0.737574  0.919126   0.270537      262  0.077332      3388  \n",
       "6  0.721805  0.735838  0.917987   0.342704      169  0.078751      2146  \n",
       "7  0.541578  0.543401  0.882353   0.226767       19  0.065744       289  \n",
       "8  0.739770  0.756811  0.939135   0.272814      319  0.061834      5159  \n",
       "9  0.678080  0.680620  0.882156   0.299429      212  0.101145      2096  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocessing = [\n",
    "    dp.BasicImputer(verbose=False),\n",
    "    dp.DataMinMaxScaler(verbose=False), \n",
    "    dp.EncoderOHE(drop=False, unknown_err=False, verbose=False)\n",
    "]\n",
    "\n",
    "dec_class = DecoupledClass(\n",
    "    cohort_col=[\"department\"],\n",
    "    transform_pipe=preprocessing,\n",
    "    estimator=get_model(),\n",
    "    minority_min_rate=0.03,\n",
    "    min_cohort_pct=0.05,\n",
    "    theta=[0.2, 0.5, 0.8],\n",
    ")\n",
    "dec_class.fit(X_train, y_train)\n",
    "\n",
    "pred = dec_class.predict_proba(X_test)\n",
    "fetch_cohort_results(X_test, y_test, pred, cohort_def=dec_class, fixed_th=True)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## DecoupledClassifier - Transfer Learning + Fairness Optimization\n",
    "\n",
    "For our final experiment, we try adding fairness optimization in our ``DecoupledClass`` object. This time, we'll optimize the Numerical Parity, which aims to make the number of positive labels for each cohort as equal as possible. We'll set $\\lambda$ to a small value to give more weight to the fairness metric. This way, we expect the estimator to sacrifice a considerable amount of performance metric (precision, recall, F1, and accuracy) over fairness (equal number of positive labels)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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>cohort</th>\n",
       "      <th>cht_query</th>\n",
       "      <th>roc</th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>threshold</th>\n",
       "      <th>num_pos</th>\n",
       "      <th>%_pos</th>\n",
       "      <th>cht_size</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>all</td>\n",
       "      <td>all</td>\n",
       "      <td>0.901364</td>\n",
       "      <td>0.885156</td>\n",
       "      <td>0.675796</td>\n",
       "      <td>0.733884</td>\n",
       "      <td>0.939001</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>607</td>\n",
       "      <td>0.036915</td>\n",
       "      <td>16443</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>cohort_0</td>\n",
       "      <td>(`department` == \"Analytics\")</td>\n",
       "      <td>0.801923</td>\n",
       "      <td>0.659641</td>\n",
       "      <td>0.613547</td>\n",
       "      <td>0.630870</td>\n",
       "      <td>0.892677</td>\n",
       "      <td>0.295972</td>\n",
       "      <td>102</td>\n",
       "      <td>0.064394</td>\n",
       "      <td>1584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cohort_1</td>\n",
       "      <td>(`department` == \"Finance\")</td>\n",
       "      <td>0.928840</td>\n",
       "      <td>0.665214</td>\n",
       "      <td>0.838909</td>\n",
       "      <td>0.695979</td>\n",
       "      <td>0.827357</td>\n",
       "      <td>0.087748</td>\n",
       "      <td>183</td>\n",
       "      <td>0.243028</td>\n",
       "      <td>753</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>cohort_2</td>\n",
       "      <td>(`department` == \"HR\")</td>\n",
       "      <td>0.910229</td>\n",
       "      <td>0.591032</td>\n",
       "      <td>0.830024</td>\n",
       "      <td>0.585841</td>\n",
       "      <td>0.764205</td>\n",
       "      <td>0.064236</td>\n",
       "      <td>200</td>\n",
       "      <td>0.284091</td>\n",
       "      <td>704</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>cohort_3</td>\n",
       "      <td>(`department` == \"Legal\")</td>\n",
       "      <td>0.917688</td>\n",
       "      <td>0.542793</td>\n",
       "      <td>0.667213</td>\n",
       "      <td>0.329452</td>\n",
       "      <td>0.373457</td>\n",
       "      <td>0.001128</td>\n",
       "      <td>222</td>\n",
       "      <td>0.685185</td>\n",
       "      <td>324</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>cohort_4</td>\n",
       "      <td>(`department` == \"Operations\")</td>\n",
       "      <td>0.894497</td>\n",
       "      <td>0.881685</td>\n",
       "      <td>0.701623</td>\n",
       "      <td>0.757745</td>\n",
       "      <td>0.938312</td>\n",
       "      <td>0.436915</td>\n",
       "      <td>155</td>\n",
       "      <td>0.045750</td>\n",
       "      <td>3388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>cohort_5</td>\n",
       "      <td>(`department` == \"Procurement\")</td>\n",
       "      <td>0.898306</td>\n",
       "      <td>0.807227</td>\n",
       "      <td>0.719237</td>\n",
       "      <td>0.753797</td>\n",
       "      <td>0.930103</td>\n",
       "      <td>0.402534</td>\n",
       "      <td>135</td>\n",
       "      <td>0.062908</td>\n",
       "      <td>2146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>cohort_6</td>\n",
       "      <td>(`department` == \"R&amp;D\")</td>\n",
       "      <td>0.699893</td>\n",
       "      <td>0.549763</td>\n",
       "      <td>0.645522</td>\n",
       "      <td>0.315951</td>\n",
       "      <td>0.342561</td>\n",
       "      <td>0.017962</td>\n",
       "      <td>211</td>\n",
       "      <td>0.730104</td>\n",
       "      <td>289</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>cohort_7</td>\n",
       "      <td>(`department` == \"Sales &amp; Marketing\")</td>\n",
       "      <td>0.938689</td>\n",
       "      <td>0.956582</td>\n",
       "      <td>0.707043</td>\n",
       "      <td>0.778220</td>\n",
       "      <td>0.956387</td>\n",
       "      <td>0.611425</td>\n",
       "      <td>162</td>\n",
       "      <td>0.031401</td>\n",
       "      <td>5159</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>cohort_8</td>\n",
       "      <td>(`department` == \"Technology\")</td>\n",
       "      <td>0.877133</td>\n",
       "      <td>0.735181</td>\n",
       "      <td>0.650417</td>\n",
       "      <td>0.679839</td>\n",
       "      <td>0.901240</td>\n",
       "      <td>0.420690</td>\n",
       "      <td>134</td>\n",
       "      <td>0.063931</td>\n",
       "      <td>2096</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     cohort                              cht_query       roc  precision  \\\n",
       "0       all                                    all  0.901364   0.885156   \n",
       "1  cohort_0          (`department` == \"Analytics\")  0.801923   0.659641   \n",
       "2  cohort_1            (`department` == \"Finance\")  0.928840   0.665214   \n",
       "3  cohort_2                 (`department` == \"HR\")  0.910229   0.591032   \n",
       "4  cohort_3              (`department` == \"Legal\")  0.917688   0.542793   \n",
       "5  cohort_4         (`department` == \"Operations\")  0.894497   0.881685   \n",
       "6  cohort_5        (`department` == \"Procurement\")  0.898306   0.807227   \n",
       "7  cohort_6                (`department` == \"R&D\")  0.699893   0.549763   \n",
       "8  cohort_7  (`department` == \"Sales & Marketing\")  0.938689   0.956582   \n",
       "9  cohort_8         (`department` == \"Technology\")  0.877133   0.735181   \n",
       "\n",
       "     recall        f1  accuracy  threshold  num_pos     %_pos  cht_size  \n",
       "0  0.675796  0.733884  0.939001   0.500000      607  0.036915     16443  \n",
       "1  0.613547  0.630870  0.892677   0.295972      102  0.064394      1584  \n",
       "2  0.838909  0.695979  0.827357   0.087748      183  0.243028       753  \n",
       "3  0.830024  0.585841  0.764205   0.064236      200  0.284091       704  \n",
       "4  0.667213  0.329452  0.373457   0.001128      222  0.685185       324  \n",
       "5  0.701623  0.757745  0.938312   0.436915      155  0.045750      3388  \n",
       "6  0.719237  0.753797  0.930103   0.402534      135  0.062908      2146  \n",
       "7  0.645522  0.315951  0.342561   0.017962      211  0.730104       289  \n",
       "8  0.707043  0.778220  0.956387   0.611425      162  0.031401      5159  \n",
       "9  0.650417  0.679839  0.901240   0.420690      134  0.063931      2096  "
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preprocessing = [\n",
    "    dp.BasicImputer(verbose=False),\n",
    "    dp.DataMinMaxScaler(verbose=False), \n",
    "    dp.EncoderOHE(drop=False, unknown_err=False, verbose=False)\n",
    "]\n",
    "\n",
    "dec_class = DecoupledClass(\n",
    "    cohort_col=[\"department\"],\n",
    "    transform_pipe=preprocessing,\n",
    "    estimator=get_model(),\n",
    "    minority_min_rate=0.03,\n",
    "    min_cohort_pct=0.05,\n",
    "    theta=[0.2, 0.5, 0.8],\n",
    "    fairness_loss=\"num_parity\",\n",
    "    lambda_coef=0.2,\n",
    "    max_joint_loss_time=600,\n",
    ")\n",
    "dec_class.fit(X_train, y_train)\n",
    "\n",
    "pred = dec_class.predict_proba(X_test)\n",
    "fetch_cohort_results(X_test, y_test, pred, cohort_def=dec_class, fixed_th=True)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see in the previous results, there was a huge drop in the performance metric of certain cohorts in order to allow the estimator to be as fair as possible according to the Numerical Parity metric. This is an approach that must be used with care, since we should always consider the nature of the dataset, and how important is the performance and fairness. We could also try different values of $\\lambda$ and try to find a better balance between performance and numerical parity."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "raipub",
   "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": "3.9.15"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "98cf402749abf383affb54f23cdde06b52ae2a6e4394659b91d1cafca4224ff4"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}