src/fair_downsampling/fair_downsampling_tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "71a98bf5-4315-4b1d-920e-09c63708c32f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# SPDX-License-Identifier: Apache-2.0\n",
    "#\n",
    "# Copyright 2024 Fujitsu Limited\n",
    "#\n",
    "# Licensed under the Apache License, Version 2.0 (the \"License\");\n",
    "# you may not use this file except in compliance with the License.\n",
    "# You may obtain a copy of the License at\n",
    "#\n",
    "#     http://www.apache.org/licenses/LICENSE-2.0\n",
    "#\n",
    "# Unless required by applicable law or agreed to in writing, software\n",
    "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
    "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
    "# See the License for the specific language governing permissions and\n",
    "# limitations under the License."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fdcfeb38-b51c-4c94-869f-9aefe2181a2b",
   "metadata": {},
   "source": [
    "## User guide of FairDownsampling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3b3650e5-bc0f-4bff-86b5-06c7be8f2ab7",
   "metadata": {},
   "source": [
    "In this tutorial we make use of the FairDownsampling algorithm.  We apply it in the Adult dataset, we compute accuracy and fairness using various common classification algorithms and we compare with the results before downsampling.  We demonstrate improvement of fairness for three common fairness metrics without sacrificing accuracy.\n",
    "\n",
    "The algorithm FairDownsampling can be applied to a wide range of imbalanced datasets.  The imbalance appears both at the class (majority vs minority) as well as the group level (e.g. gender, age, race).  In traditional downsampling methods, such as NearMiss, there is usually a trade-off between accuracy and fairness. Downsampling would often slightly increase accuracy, but in most cases fairness deteriorates significantly. The proposed downsampling algorithm considers both fairness and accuracy, while trying to avoid underfitting, by deleting samples away from the class and group boundaries.\n",
    "\n",
    "We demonstrate our algorithm using the Adult dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e838be3e-2df5-4442-affe-6b65e964df6d",
   "metadata": {},
   "source": [
    "#### imports"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99b0f108-f436-4279-b520-53e923a006d7",
   "metadata": {},
   "source": [
    "Necessary imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "9bc81215-915a-4db7-b0e3-06a8d2d22bd2",
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.metrics import balanced_accuracy_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "from xgboost import XGBClassifier\n",
    "from imblearn.under_sampling import NearMiss #controlled under-sampling algorithm\n",
    "from ucimlrepo import fetch_ucirepo #for downloading datasets\n",
    "\n",
    "import aif360\n",
    "from aif360 import metrics\n",
    "\n",
    "import fairdownsampling as fd"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5008616f-a1dd-4dfa-82f9-346fa2c0f63e",
   "metadata": {},
   "source": [
    "#### Use case guide: Adult dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "78de41ba-2f8d-4900-8722-07f25f90cab8",
   "metadata": {},
   "source": [
    "We use the common dataset Adult which can be downloaded from <br> https://archive.ics.uci.edu/dataset/2/adult<br>\n",
    "We then prepare it and convert it to aif360 format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e79efed4-5e74-4825-ab56-78ddd495cfc9",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "adult = fetch_ucirepo(id=2) \n",
    "# input data (as pandas dataframes) \n",
    "X_adult = adult.data.features \n",
    "y_adult = adult.data.targets\n",
    "#preprocess and prepare Adult dataset\n",
    "adult_total = pd.concat([X_adult, y_adult], axis=1)\n",
    "adult_total.replace({'income': {'<=50K.': '<=50K', '>50K.': '>50K'}}, inplace=True)\n",
    "adult1 = adult_total.dropna()\n",
    "adult2 = adult1.drop_duplicates()\n",
    "for col in list(adult2.columns):\n",
    "    adult2[[col]] = LabelEncoder().fit_transform(adult2[[col]].to_numpy().reshape(-1)).reshape(-1,1)\n",
    "adult_final = adult2.drop(columns=['fnlwgt']).drop_duplicates()\n",
    "X_adult_final = adult_final.drop(columns=['income'])\n",
    "y_adult_final = adult_final['income']\n",
    "\n",
    "#convert Adult dataset into AIF360 dataset format\n",
    "dataset_adult = X_adult_final.copy()\n",
    "dataset_adult[\"income\"] = y_adult_final.copy()\n",
    "aif360_adult = aif360.datasets.BinaryLabelDataset(\n",
    "    df=dataset_adult,\n",
    "    label_names=[\"income\"],\n",
    "    favorable_label=1.0,\n",
    "    unfavorable_label=0.0,\n",
    "    protected_attribute_names=[\"sex\"],\n",
    "    unprivileged_protected_attributes=[np.array([0.0])],\n",
    "    privileged_protected_attributes=[np.array([1.0])])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "b5f50af2-863a-4544-b2c0-e66275ac2c8e",
   "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>age</th>\n",
       "      <th>workclass</th>\n",
       "      <th>education</th>\n",
       "      <th>education-num</th>\n",
       "      <th>marital-status</th>\n",
       "      <th>occupation</th>\n",
       "      <th>relationship</th>\n",
       "      <th>race</th>\n",
       "      <th>sex</th>\n",
       "      <th>capital-gain</th>\n",
       "      <th>capital-loss</th>\n",
       "      <th>hours-per-week</th>\n",
       "      <th>native-country</th>\n",
       "      <th>income</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>22</td>\n",
       "      <td>7</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>33</td>\n",
       "      <td>6</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>12</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>21</td>\n",
       "      <td>4</td>\n",
       "      <td>11</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>36</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>11</td>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>5</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",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48835</th>\n",
       "      <td>31</td>\n",
       "      <td>2</td>\n",
       "      <td>12</td>\n",
       "      <td>13</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48836</th>\n",
       "      <td>16</td>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>4</td>\n",
       "      <td>10</td>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48837</th>\n",
       "      <td>22</td>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>0</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>35</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48840</th>\n",
       "      <td>27</td>\n",
       "      <td>4</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>82</td>\n",
       "      <td>0</td>\n",
       "      <td>39</td>\n",
       "      <td>39</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48841</th>\n",
       "      <td>18</td>\n",
       "      <td>5</td>\n",
       "      <td>9</td>\n",
       "      <td>12</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>59</td>\n",
       "      <td>39</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>41432 rows × 14 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       age  workclass  education  education-num  marital-status  occupation  \\\n",
       "0       22          7          9             12               4           1   \n",
       "1       33          6          9             12               2           4   \n",
       "2       21          4         11              8               0           6   \n",
       "3       36          4          1              6               2           6   \n",
       "4       11          4          9             12               2          10   \n",
       "...    ...        ...        ...            ...             ...         ...   \n",
       "48835   31          2         12             13               0           8   \n",
       "48836   16          4          9             12               4          10   \n",
       "48837   22          4          9             12               0          10   \n",
       "48840   27          4          9             12               0           1   \n",
       "48841   18          5          9             12               2           4   \n",
       "\n",
       "       relationship  race  sex  capital-gain  capital-loss  hours-per-week  \\\n",
       "0                 1     4    1            27             0              39   \n",
       "1                 0     4    1             0             0              12   \n",
       "2                 1     4    1             0             0              39   \n",
       "3                 0     2    1             0             0              39   \n",
       "4                 5     2    0             0             0              39   \n",
       "...             ...   ...  ...           ...           ...             ...   \n",
       "48835             1     4    1             0             0              39   \n",
       "48836             3     4    1             0             0              39   \n",
       "48837             1     4    0             0             0              35   \n",
       "48840             3     1    1            82             0              39   \n",
       "48841             0     4    1             0             0              59   \n",
       "\n",
       "       native-country  income  \n",
       "0                  39       0  \n",
       "1                  39       0  \n",
       "2                  39       0  \n",
       "3                  39       0  \n",
       "4                   5       0  \n",
       "...               ...     ...  \n",
       "48835              39       0  \n",
       "48836              39       0  \n",
       "48837              39       0  \n",
       "48840              39       0  \n",
       "48841              39       1  \n",
       "\n",
       "[41432 rows x 14 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#take a look at the dataset\n",
    "adult_final"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "813631be-011f-4c00-ba49-518e00015b70",
   "metadata": {},
   "source": [
    "#### Accuracy and Fairness before FairDownsampling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4eee7e42-5a6a-48d2-92bd-c4465a7b30dd",
   "metadata": {},
   "source": [
    "We calculate accuracy and fairness before downsampling for seven common classification algorithms.  We store the results in a list for later comparison.\n",
    "\n",
    "We measure accuracy performance using the Balanced Accuracy metric, and we measure fairness using Statistical Parity Difference, Equal Opportunity Difference, and Average Odds Difference."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "203e49cc-91ce-413c-918e-6dfba5f34bd3",
   "metadata": {},
   "outputs": [],
   "source": [
    "models = [LogisticRegression(max_iter=10000),\n",
    "          RandomForestClassifier(),\n",
    "          KNeighborsClassifier(),\n",
    "          XGBClassifier(),\n",
    "          GaussianNB(),\n",
    "          GradientBoostingClassifier(),\n",
    "          DecisionTreeClassifier()\n",
    "         ]\n",
    "\n",
    "metric_1_before = [] #balanced accuracy\n",
    "metric_2_before = [] #statistical parity difference\n",
    "metric_3_before = [] #equal opportunity difference\n",
    "metric_4_before = [] #average odds difference\n",
    "    \n",
    "for model in models:\n",
    "    \n",
    "    metric_1_temp = []\n",
    "    metric_2_temp = []\n",
    "    metric_3_temp = []\n",
    "    metric_4_temp = []\n",
    "    \n",
    "    for random_state in [40, 41, 42, 43, 44, 45, 46, 47, 48, 49]:\n",
    "\n",
    "        X_train, X_test, y_train, y_test = train_test_split(X_adult_final, y_adult_final, test_size=0.30, random_state=random_state)\n",
    "        \n",
    "        model.fit(X_train.values, y_train.values)\n",
    "        y_predicted = model.predict(X_test.values)\n",
    "\n",
    "        dataset_before = X_test.copy()\n",
    "        dataset_before[\"Labels\"] = y_test.copy()\n",
    "        aif360_before = aif360.datasets.BinaryLabelDataset(\n",
    "            df=dataset_before,\n",
    "            label_names=[\"Labels\"],\n",
    "            protected_attribute_names=[\"sex\"])\n",
    "\n",
    "        dataset_after = X_test.copy()\n",
    "        dataset_after[\"Labels\"] = y_predicted.copy()\n",
    "        aif360_after = aif360.datasets.BinaryLabelDataset(\n",
    "            df=dataset_after,\n",
    "            label_names=[\"Labels\"],\n",
    "            protected_attribute_names=[\"sex\"])\n",
    "\n",
    "        classification_metric = aif360.metrics.ClassificationMetric(\n",
    "            dataset=aif360_before,\n",
    "            classified_dataset=aif360_after,\n",
    "            unprivileged_groups=[{\"sex\": 0}],\n",
    "            privileged_groups=[{\"sex\": 1}])\n",
    "\n",
    "        metric_1_temp.append(np.abs(balanced_accuracy_score(y_test, y_predicted)))\n",
    "        metric_2_temp.append(-(classification_metric.statistical_parity_difference()))\n",
    "        metric_3_temp.append(-(classification_metric.equal_opportunity_difference()))\n",
    "        metric_4_temp.append(-(classification_metric.average_odds_difference()))\n",
    "\n",
    "    metric_1_before.append(np.mean(metric_1_temp))\n",
    "    metric_2_before.append(np.mean(metric_2_temp))\n",
    "    metric_3_before.append(np.mean(metric_3_temp))\n",
    "    metric_4_before.append(np.mean(metric_4_temp))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "83ae66c2-fce1-4fc9-955e-a94e3a737f95",
   "metadata": {},
   "source": [
    "#### FairDownsampling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e160202-742a-4313-94a0-4d67cab0c8d9",
   "metadata": {},
   "source": [
    "We next perform FairDownsampling on the Adult dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "0966920c-3dce-44f1-b123-8bcd02e39a5d",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_adult = fd.FairDownSampler(sampling_strategy=0.8)\n",
    "downsampled_adult = model_adult.fit_resample(aif360_adult)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59e238c2-c0eb-4cd0-beb7-1cd760c7180a",
   "metadata": {},
   "source": [
    "#### Accuracy and Fairness after FairDownsampling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cd8345b-25ce-4234-82cc-53dba1b61477",
   "metadata": {},
   "source": [
    "We then compute Accuracy and Fairness after FairDownsampling.  We store the results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "8e509b86-916d-4f21-9b03-607017f9126b",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_resampled = ((downsampled_adult.convert_to_dataframe())[0]).drop(columns=['income'])\n",
    "y_resampled = ((downsampled_adult.convert_to_dataframe())[0])['income']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9ff4bf6b-5cd7-44f1-a975-c2135d8caa67",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "models = [LogisticRegression(max_iter=10000),\n",
    "          RandomForestClassifier(),\n",
    "          KNeighborsClassifier(),\n",
    "          XGBClassifier(),\n",
    "          GaussianNB(),\n",
    "          GradientBoostingClassifier(),\n",
    "          DecisionTreeClassifier()\n",
    "         ]\n",
    "\n",
    "metric_1_after = [] #balanced accuracy\n",
    "metric_2_after = [] #statistical parity difference\n",
    "metric_3_after = [] #equal opportunity difference\n",
    "metric_4_after = [] #average odds difference\n",
    "    \n",
    "for model in models:\n",
    "    \n",
    "    metric_1_temp = []\n",
    "    metric_2_temp = []\n",
    "    metric_3_temp = []\n",
    "    metric_4_temp = []\n",
    "    \n",
    "    for random_state in [40, 41, 42, 43, 44, 45, 46, 47, 48, 49]:\n",
    "\n",
    "        X_train, X_test, y_train, y_test = train_test_split(X_resampled, y_resampled, test_size=0.30, random_state=random_state)\n",
    "        \n",
    "        model.fit(X_train, y_train)\n",
    "        y_predicted = model.predict(X_test.values)\n",
    "\n",
    "        dataset_before = X_test.copy()\n",
    "        dataset_before[\"Labels\"] = y_test.copy()\n",
    "        aif360_before = aif360.datasets.BinaryLabelDataset(\n",
    "            df=dataset_before,\n",
    "            label_names=[\"Labels\"],\n",
    "            protected_attribute_names=[\"sex\"])\n",
    "\n",
    "        dataset_after = X_test.copy()\n",
    "        dataset_after[\"Labels\"] = y_predicted.copy()\n",
    "        aif360_after = aif360.datasets.BinaryLabelDataset(\n",
    "            df=dataset_after,\n",
    "            label_names=[\"Labels\"],\n",
    "            protected_attribute_names=[\"sex\"])\n",
    "\n",
    "        classification_metric = aif360.metrics.ClassificationMetric(\n",
    "            dataset=aif360_before,\n",
    "            classified_dataset=aif360_after,\n",
    "            unprivileged_groups=[{\"sex\": 0}],\n",
    "            privileged_groups=[{\"sex\": 1}])\n",
    "\n",
    "        metric_1_temp.append(balanced_accuracy_score(y_test, y_predicted))\n",
    "        metric_2_temp.append(-classification_metric.statistical_parity_difference())\n",
    "        metric_3_temp.append(-classification_metric.equal_opportunity_difference())\n",
    "        metric_4_temp.append(-classification_metric.average_odds_difference())\n",
    "\n",
    "    metric_1_after.append(np.mean(metric_1_temp))\n",
    "    metric_2_after.append(np.mean(metric_2_temp))\n",
    "    metric_3_after.append(np.mean(metric_3_temp))\n",
    "    metric_4_after.append(np.mean(metric_4_temp))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "209837ad-e0cc-4ac9-b090-71dfd09fb49a",
   "metadata": {},
   "source": [
    "#### Comparison before and after FairDownsampling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d39ad759-63b6-4f34-ac7b-f5a9cee14e24",
   "metadata": {},
   "source": [
    "We want to compare our accuracy and fairness before and after downsampling.  Thus, we plot the obtained results as accuracy vs fairness for the three different fairness metrics.  The plots demonstrate improvements in fairness without sacrificing accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "52260ec5-264c-4ef7-ac47-556f912685f9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x400 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "markersize = 5\n",
    "fig, ax = plt.subplots(nrows=1, ncols=3, figsize=(12, 4));\n",
    "fig.tight_layout(pad=4.0)\n",
    "ax[0].set_title(\"accuracy vs statistical parity\")\n",
    "ax[0].set_xlabel(\"accuracy\")\n",
    "ax[0].set_ylabel(\"statistical parity\")\n",
    "ax[1].set_title(\"accuracy vs equal opportunity\")\n",
    "ax[1].set_xlabel(\"accuracy\")\n",
    "ax[1].set_ylabel(\"equal opportunity\")\n",
    "ax[2].set_title(\"accuracy vs odds difference\")\n",
    "ax[2].set_xlabel(\"accuracy\")\n",
    "ax[2].set_ylabel(\"odds difference\")\n",
    "x = np.linspace(0.5, 1, 100)\n",
    "y = np.linspace(0, 0, 100)\n",
    "ax[0].plot(x, y, linestyle='solid', c='gray')\n",
    "ax[1].plot(x, y, linestyle='solid', c='gray')\n",
    "ax[2].plot(x, y, linestyle='solid', c='gray')\n",
    "ax[0].plot(metric_1_before, metric_2_before, 'o', label='before downsampling', markersize=markersize);\n",
    "ax[1].plot(metric_1_before, metric_3_before, 'o', label='before downsampling', markersize=markersize);\n",
    "ax[2].plot(metric_1_before, metric_4_before, 'o', label='before downsampling', markersize=markersize);\n",
    "ax[0].plot(metric_1_after, metric_2_after, 'o', label='after fair downsampling', markersize=markersize);\n",
    "ax[1].plot(metric_1_after, metric_3_after, 'o', label='after fair downsampling', markersize=markersize);\n",
    "ax[2].plot(metric_1_after, metric_4_after, 'o', label='after fair downsampling', markersize=markersize);\n",
    "ax[0].set_xlim([0.6, 0.9])\n",
    "ax[1].set_xlim([0.6, 0.9])\n",
    "ax[2].set_xlim([0.6, 0.9])\n",
    "ax[0].set_ylim([-0.1, 0.4])\n",
    "ax[1].set_ylim([-0.1, 0.35])\n",
    "ax[2].set_ylim([-0.1, 0.25])\n",
    "ax[0].legend(loc=\"upper left\")\n",
    "ax[1].legend(loc=\"upper left\")\n",
    "ax[2].legend(loc=\"upper left\");\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.12"
  },
  "toc-autonumbering": false,
  "toc-showcode": false,
  "toc-showmarkdowntxt": false,
  "toc-showtags": false
 },
 "nbformat": 4,
 "nbformat_minor": 5
}