SimpleDebiasing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.compose import make_column_selector as selector\n",
    "from sklearn.preprocessing import OneHotEncoder, StandardScaler\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('us_census.csv', sep=\",\")\n",
    "target = data['income']\n",
    "data = data.drop(columns=['income'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_train, data_test, target_train, target_test = train_test_split(\n",
    "    data, target, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<=50K    0.761841\n",
       ">50K     0.238159\n",
       "Name: income, dtype: float64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target_train.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocessing pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "numerical_columns = selector(dtype_exclude=object)(data)\n",
    "categorical_columns = selector(dtype_include=object)(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "categorical_preprocessor = OneHotEncoder(handle_unknown=\"ignore\")\n",
    "numerical_preprocessor = StandardScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "preprocessor = ColumnTransformer([\n",
    "    ('one-hot-encoder', categorical_preprocessor, categorical_columns),\n",
    "    ('standard_scaler', numerical_preprocessor, numerical_columns)])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create the fair model wrapper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "class fairModelWrapper:\n",
    "    \n",
    "    def __init__(self, model):\n",
    "        self.model = model\n",
    "\n",
    "    def fit(self, data, target, plot=True):\n",
    "        data_target = data.copy()\n",
    "        data_target['predictions'] = self.model.predict(data)\n",
    "        \n",
    "        income_proba = data_target['predictions'].value_counts(normalize=True)['<=50K']\n",
    "        \n",
    "        female_predicted_probas = self.model.predict_proba(data_target[data_target['sex']=='Female'])[:,1]\n",
    "        self.female_threshold = np.quantile(female_predicted_probas, income_proba)\n",
    "        \n",
    "        male_predicted_probas = self.model.predict_proba(data_target[data_target['sex']=='Male'])[:,1]\n",
    "        self.male_threshold = np.quantile(male_predicted_probas, income_proba)\n",
    "        \n",
    "        if plot:\n",
    "            df_probas_female = pd.DataFrame(female_predicted_probas, columns=['probas'])\n",
    "            df_probas_female['sex'] = 'female'\n",
    "\n",
    "            df_probas_male = pd.DataFrame(male_predicted_probas, columns=['probas'])\n",
    "            df_probas_male['sex'] = 'male'\n",
    "\n",
    "            df_probas = pd.concat((df_probas_female,df_probas_male), ignore_index=True)\n",
    "\n",
    "            sns.histplot(data=df_probas, x='probas', hue='sex', bins=20, kde=True, palette='viridis', stat='percent')\n",
    "            plt.axvline(self.female_threshold, 0,0.7, c='tab:blue')\n",
    "            plt.axvline(self.male_threshold, 0,0.7, c='mediumseagreen')\n",
    "            plt.xlabel('Predicted probability of having an income above 50K')\n",
    "            plt.annotate('Threshold for males to \\nbe predicted as high income',\n",
    "                    xy =(0.63, 10), color='tab:green'\n",
    "                )\n",
    "            plt.annotate('Threshold for females to \\nbe predicted as high income',\n",
    "                    xy =(0., 23), color='tab:blue'\n",
    "                )\n",
    "            plt.ylim(0,30)\n",
    "            plt.xlim(-0.05,1.3)\n",
    "            plt.savefig('images/threshold.png')\n",
    "        \n",
    "    \n",
    "    def predict(self, data_test):\n",
    "        data_target = data_test.copy()\n",
    "        data_target['predictions'] = self.model.predict(data_test)\n",
    "        \n",
    "        female_predicted_probas = self.model.predict_proba(data_target[data_target['sex']=='Female'])[:,1]\n",
    "        female_predictions_fair = np.where(female_predicted_probas>self.female_threshold, '>50K', '<=50K')\n",
    "        \n",
    "        male_predicted_probas = self.model.predict_proba(data_target[data_target['sex']=='Male'])[:,1]\n",
    "        male_predictions_fair = np.where(male_predicted_probas>self.male_threshold, '>50K', '<=50K')\n",
    "        \n",
    "        data_target.loc[data_target['sex']=='Female', 'predictions'] = female_predictions_fair        \n",
    "        data_target.loc[data_target['sex']=='Male', 'predictions'] = male_predictions_fair        \n",
    "        return np.array(data_target['predictions'])\n",
    "        \n",
    "        \n",
    "    def score(self, data, target):\n",
    "        predictions = self.predict(data)\n",
    "        return accuracy_score(predictions, target)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit the model and wrap it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = make_pipeline(preprocessor, LogisticRegression(max_iter=500))\n",
    "model.fit(data_train, target_train)\n",
    "fairmodel = fairModelWrapper(model)\n",
    "fairmodel.fit(data_train, target_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate fairness and accuracy of the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_scores_and_plot_disparities(data_test, target_test, model):\n",
    "    predictions_test = model.predict(data_test)\n",
    "    \n",
    "    tmp = data_test.copy()\n",
    "    tmp['Model predictions'] = predictions_test\n",
    "    tmp['Data'] = target_test\n",
    "    \n",
    "    grouped_predictions = tmp.groupby('sex')['Model predictions'].value_counts(normalize=True)\n",
    "    dp_model = np.abs(grouped_predictions['Female']['>50K'] - grouped_predictions['Male']['>50K'])\n",
    "    \n",
    "    grouped_gt = tmp.groupby('sex')['Data'].value_counts(normalize=True)\n",
    "    dp_data = np.abs(grouped_gt['Female']['>50K'] - grouped_gt['Male']['>50K'])\n",
    "    \n",
    "    fig, ax = plt.subplots(figsize=(6, 2))\n",
    "    arguments = dict(data=tmp, x='Model predictions', hue='sex', kde=False, multiple=\"dodge\", stat=\"percent\", common_norm=False, discrete=True, shrink=.3, ax=ax, palette='viridis')\n",
    "    sns.histplot(**arguments)\n",
    "    ax.legend(handles=ax.legend_.legendHandles, labels=[t.get_text() for t in ax.legend_.texts],\n",
    "              title=ax.legend_.get_title().get_text(),\n",
    "              bbox_to_anchor=(1.05, 1), loc='upper left')\n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "    fig, ax = plt.subplots(figsize=(6, 2))\n",
    "    arguments = dict(data=tmp, x='Data', hue='sex', kde=False, multiple=\"dodge\", stat=\"percent\", common_norm=False, discrete=True, shrink=.3, ax=ax, palette='viridis')\n",
    "    p = sns.histplot(**arguments)\n",
    "    ax.legend(handles=ax.legend_.legendHandles, labels=[t.get_text() for t in ax.legend_.texts],\n",
    "              title=ax.legend_.get_title().get_text(),\n",
    "              bbox_to_anchor=(1.05, 1), loc='upper left')\n",
    "    plt.show()    \n",
    "    \n",
    "    print('\\n','dp_model=%s\\n'%dp_model, 'dp_data=%s\\n'%dp_data, 'accuracy=%s'%model.score(data_test, target_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " dp_model=0.175768072280926\n",
      " dp_data=0.1920244613837473\n",
      " accuracy=0.8466956023257718\n"
     ]
    }
   ],
   "source": [
    "compute_scores_and_plot_disparities(data_test, target_test, model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " dp_model=0.0019138844306005598\n",
      " dp_data=0.1920244613837473\n",
      " accuracy=0.8263860453689297\n"
     ]
    }
   ],
   "source": [
    "compute_scores_and_plot_disparities(data_test, target_test, fairmodel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "last_runtime": {
    "build_target": "//learning/deepmind/public/tools/ml_python:ml_notebook",
    "kind": "private"
   },
   "name": "OTT fairness: tutorial",
   "provenance": [
    {
     "file_id": "1PYCJcF0LS09dQVdwI73b2ZJbQesYv3zu",
     "timestamp": 1622585910089
    }
   ],
   "toc_visible": true
  },
  "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}