LR-Tuning-Controllable.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn import linear_model, metrics\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('datasets/controllable_sampled.csv', index_col = 'id')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building a baseline LR model using all \"controllable\" features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train0, X_test0, Y_train0, Y_test0 = train_test_split(df.drop('Label', 1), df.Label, test_size = 0.2, random_state = 42)\n",
    "\n",
    "scaler = StandardScaler().fit(X_train0)\n",
    "X_train_scaled0 = pd.DataFrame(scaler.transform(X_train0), columns = df.drop('Label', 1).columns)\n",
    "X_test_scaled0 = pd.DataFrame(scaler.transform(X_test0), columns = df.drop('Label', 1).columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr0 = linear_model.LogisticRegression(max_iter = 2000, solver = 'saga')\n",
    "lr0.fit(X_train_scaled0, Y_train0)\n",
    "Y_pred = lr0.predict_proba(X_test_scaled0)[:,1]\n",
    "\n",
    "fpr_lr0, tpr_lr0, thresholds_lr0 = metrics.roc_curve(Y_test0, Y_pred)\n",
    "auc_lr0 = metrics.auc(fpr_lr0, tpr_lr0)\n",
    "acc_lr0 = np.mean(np.abs(Y_test0 - Y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize = (12,7))\n",
    "\n",
    "plt.plot(fpr_lr0, tpr_lr0, color = \"darkseagreen\", label = \"LR (AUC = {:0.4f})\".format(auc_lr0))\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.xlabel('FPR')\n",
    "plt.ylabel('TPR')\n",
    "plt.title('ROC')\n",
    "plt.legend(loc = 'lower right')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Baseline LR model accuracy = 0.4432\n"
     ]
    }
   ],
   "source": [
    "print('Baseline LR model accuracy = {:.4f}'.format(acc_lr0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, estimator=LogisticRegression(), n_jobs=-1,\n",
       "                   param_distributions=[{'C': array([1.00000000e-05, 1.02331658e-05, 1.04717682e-05, 1.07159340e-05,\n",
       "       1.09657929e-05, 1.12214777e-05, 1.14831241e-05, 1.17508713e-05,\n",
       "       1.20248614e-05, 1.23052400e-05, 1.25921561e-05, 1.28857621e-05,\n",
       "       1.31862140e-05, 1.34936714e-05, 1.38082977e-05, 1.41302599e-05,\n",
       "       1.44597292e-...\n",
       "       7.07701066e+04, 7.24202233e+04, 7.41088152e+04, 7.58367791e+04,\n",
       "       7.76050334e+04, 7.94145172e+04, 8.12661920e+04, 8.31610415e+04,\n",
       "       8.51000725e+04, 8.70843150e+04, 8.91148232e+04, 9.11926760e+04,\n",
       "       9.33189772e+04, 9.54948564e+04, 9.77214697e+04, 1.00000000e+05]),\n",
       "                                         'max_iter': [2000],\n",
       "                                         'penalty': ['l1', 'l2'],\n",
       "                                         'solver': ['saga']}],\n",
       "                   random_state=42, refit='roc_auc',\n",
       "                   scoring=['roc_auc', 'accuracy'])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "\n",
    "# Hyperparameters over which to search\n",
    "param_grid = [\n",
    "    {'C': np.logspace(-5, 5, num = 1000), \n",
    "     'penalty': ['l1', 'l2'],\n",
    "    'solver': ['saga'],\n",
    "    'max_iter': [2000]}\n",
    "]\n",
    "\n",
    "scores = ['roc_auc', 'accuracy']\n",
    "\n",
    "lr_RS = RandomizedSearchCV(linear_model.LogisticRegression(), param_distributions = param_grid, scoring = scores, \n",
    "                     cv = 5, refit = scores[0], n_jobs = -1, random_state = 42)\n",
    "lr_RS.fit(X_train_scaled0, Y_train0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'solver': 'saga', 'penalty': 'l1', 'max_iter': 2000, 'C': 0.20151357338155584}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr_RS.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC of best performing model: 0.68852\n"
     ]
    }
   ],
   "source": [
    "print(\"AUC of best performing model: {:.5f}\".format(lr_RS.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}