4_ML_Prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2ws3c9OS89Cn"
   },
   "source": [
    "\n",
    "\n",
    "---\n",
    "\n",
    "\n",
    "# **ML prediction**\n",
    "#### ***Look before you Leap : Leveraging Predictive Models to Improve Automotive Safety and Travel Time***\n",
    "\n",
    "\n",
    "\n",
    "---\n",
    "\n",
    "\n",
    "\n",
    "**Final Project**\n",
    "CSCI 35300/79502  \n",
    "##### Hannah Do\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "### Comparison on datasets with addition of twitter features \n",
    "- importing different files for cross validation and evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1373,
     "status": "ok",
     "timestamp": 1607165536120,
     "user": {
      "displayName": "Hannah Do",
      "photoUrl": "",
      "userId": "09169567944837111458"
     },
     "user_tz": -540
    },
    "id": "cFyz34eXc8gy",
    "outputId": "46810445-a3f3-450f-a612-74338826634f"
   },
   "outputs": [],
   "source": [
    "# Import all the libraries\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import seaborn as sns\n",
    "import csv\n",
    "import datetime, string, re\n",
    "import sys, os, warnings\n",
    "\n",
    "from numpy import sqrt, argmax\n",
    "from collections import Counter\n",
    "\n",
    "from scipy import stats\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "from imblearn.over_sampling import RandomOverSampler\n",
    "from imblearn.ensemble import BalancedRandomForestClassifier\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.model_selection import cross_val_score, cross_val_predict\n",
    "from sklearn.feature_selection import SelectFromModel\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import roc_curve, auc, roc_auc_score, precision_recall_curve, confusion_matrix, precision_score, recall_score, f1_score\n",
    "from sklearn.metrics import make_scorer\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "**Different ML models**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a function with many machine learning models \n",
    "\n",
    "def models_cross(X,y):\n",
    "\n",
    "    list1 = []\n",
    "\n",
    "    # kNN\n",
    "    knn = KNeighborsClassifier()\n",
    "    knn.fit(X_train, y_train)\n",
    "    y_pred=knn.predict(X_test)\n",
    "\n",
    "    knn_acc=accuracy_score(y_test, y_pred)\n",
    "    list1.append(knn_acc)\n",
    "    print('kNN accuracy score : ', knn_acc)\n",
    "    \n",
    "\n",
    "   # SVC (poly kernel)\n",
    "    svc = SVC(random_state = 4, C= 50, gamma= 'scale', kernel= 'poly')\n",
    "    svc_scores = cross_val_score(svc, X, y, cv=8,n_jobs=-1)\n",
    "#     print('SVC Cross-Validation Accuracy Scores', svc_scores)\n",
    "    svc_scores_series = pd.Series(svc_scores)\n",
    "    list1.append(svc_scores_series.mean())\n",
    "    print('SVC accuracy score : ', svc_scores_series.mean())\n",
    "\n",
    "\n",
    "    # GaussianNB\n",
    "    gauss = GaussianNB()\n",
    "    gauss_scores = cross_val_score(gauss, X, y, cv=8, n_jobs=-1)\n",
    "#     print('GaussianNB Cross-Validation Accuracy Scores', gauss_scores)\n",
    "    gauss_scores_series = pd.Series(gauss_scores)\n",
    "    list1.append(gauss_scores_series.mean())\n",
    "    print('Gaussian NB accuracy score : ', gauss_scores_series.mean())\n",
    "\n",
    "            \n",
    "    # Decision Tree\n",
    "    tree = DecisionTreeClassifier(criterion = 'gini', max_depth='None', max_features='None', min_samples_leaf=2, min_samples_split=3)\n",
    "    tree_scores = cross_val_score(tree, X, y, cv=8,n_jobs=-1)\n",
    "#     print('Decision Tree Cross-Validation Accuracy Scores', tree_scores)\n",
    "    tree_scores_series = pd.Series(tree_scores)\n",
    "    list1.append(tree_scores_series.mean())\n",
    "    print('Decision Tree accuracy score : ', tree_scores_series.mean())\n",
    "\n",
    "\n",
    "    # Random Forest Classifier \n",
    "    forest = RandomForestClassifier(n_estimators=1000, max_features = 'sqrt')\n",
    "    forest_scores = cross_val_score(forest, X, y, cv=8,n_jobs=-1)\n",
    "#     print('Random forest Cross-Validation Accuracy Scores', forest_scores)\n",
    "    forest_scores_series = pd.Series(forest_scores)\n",
    "    list1.append(forest_scores_series.mean())\n",
    "    print('Random Forest accuracy score : ', forest_scores_series.mean())\n",
    "   \n",
    "\n",
    "    # Logistic Regression\n",
    "    # multi_class='ovr', c=0.1, penalty='l2', solver='newton-cg'\n",
    "    log = LogisticRegression()\n",
    "    log_scores = cross_val_score(log, X, y, cv=8,n_jobs=-1)\n",
    "#     print('Logistic regression Cross-Validation Accuracy Scores', log_scores)\n",
    "    log_scores_series = pd.Series(log_scores)\n",
    "    list1.append(log_scores_series.mean())\n",
    "    print('Logistic Regression accuracy score : ', log_scores_series.mean())\n",
    "\n",
    "    \n",
    "    # XGBoost Classifier\n",
    "    xg = XGBClassifier(objective='binary:logistic', learning_rate=0.5, max_depth=5)\n",
    "    xg_scores = cross_val_score(xg, X, y, cv=8,n_jobs=-1)\n",
    "#     print('XGBoost Cross-Validation Accuracy Scores', xg_scores)\n",
    "    xg_scores_series = pd.Series(xg_scores)\n",
    "    list1.append(xg_scores_series.mean())\n",
    "    print('XGBoost accuracy score : ', xg_scores_series.mean())\n",
    "\n",
    "    return list1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "rjPC3qZKc8gz",
    "outputId": "50f28bad-4862-4acb-b58d-40d4b36a6815",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "final_table = pd.DataFrame()\n",
    "\n",
    "files = ['processed_1mo_smote', 'processed_1mo_realistic', 'processed_3mo_smote', 'processed_3mo_realistic', 'processed_6mo_realistic_short', 'processed_6mo_smote_short']\n",
    "# processed_1mo_smote\n",
    "# processed_1mo_realistic\n",
    "# processed_3mo_smote\n",
    "# ...\n",
    "\n",
    "i = 1\n",
    "    \n",
    "for f in files:\n",
    "\n",
    "    print('\\n' + f)\n",
    "    temp = pd.read_csv('merged/' + f + '.csv', sep=',')\n",
    "\n",
    "    temp = temp.drop(columns = ['Unnamed: 0'], axis=1)\n",
    "    # drop NaN column with index values\n",
    "\n",
    "    list1 = pd.DataFrame()\n",
    "\n",
    "    y = temp['target']\n",
    "    X = temp.drop('target', axis=1)\n",
    "\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)\n",
    "\n",
    "    list1 = models_cross(X,y)\n",
    "\n",
    "    final_table['twitter '+ f + str(i)] = list1\n",
    "    \n",
    "    \n",
    "    print('\\n' + f + ' accident record')\n",
    "\n",
    "    list2 = pd.DataFrame()\n",
    "\n",
    "    y = temp['target']\n",
    "    X = temp.drop('target', axis=1)\n",
    "\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)\n",
    "\n",
    "    list2 = models_cross(X,y)\n",
    "\n",
    "    final_table['accident '+ f + str(i)] = list2\n",
    "\n",
    "    i = i+1\n",
    "    \n",
    "final_table\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "jyyYHwslc8g0",
    "outputId": "a5adf8a5-91ff-4556-cf9d-a18aa782f75b",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "ml_models = ['KNN', 'SVC', 'Gaussian NB', 'Decision Tree', 'Random Forest', 'Logistic Regression', 'XGBoost']\n",
    "# 'SVC-rbf', \n",
    "\n",
    "final_table.index = ml_models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature Importance - Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp = pd.read_csv('02032022/smote_1mo.csv', sep=',')\n",
    "\n",
    "temp = temp.drop(columns = ['Unnamed: 0'], axis=1)\n",
    "# drop NaN column with index values\n",
    "\n",
    "list0 = pd.DataFrame()\n",
    "\n",
    "y = temp['target']\n",
    "X = temp.drop('target', axis=1)    \n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "forest = RandomForestClassifier(n_estimators=10, criterion='entropy', random_state=0)\n",
    "forest.fit(X_train, y_train)\n",
    "\n",
    "feature_imp = pd.Series(forest.feature_importances_,index=X_train.columns).sort_values(ascending=False)\n",
    "\n",
    "k=10\n",
    "sns.barplot(x=feature_imp[:10], y=feature_imp.index[:k])\n",
    "plt.xlabel('Feature Importance Score')\n",
    "plt.ylabel('Features')\n",
    "plt.title(\"Visualizing Important Features\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation on Random Forest Model\n",
    "1. Dataset with Accidents, Smote method, and boosting Negative Instances - 6 months"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp = pd.read_csv('02032022/negboost_1mo.csv', sep=',')\n",
    "\n",
    "temp = temp.drop(columns = ['Unnamed: 0'], axis=1)\n",
    "# drop NaN column with index values\n",
    "\n",
    "y = temp['target']\n",
    "X = temp.drop('target', axis=1)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "y_pred = cross_val_predict(forest, X_test, y_test, method=\"predict_proba\", cv=5)[:, 1]\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_test, y_pred)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal Threshold:  0.8\n",
      "Coordinate : ( 0.004701995170923879 ,  0.9962985811227637 )\n",
      "AUC score :  0.9998857368997847\n"
     ]
    }
   ],
   "source": [
    "geomean_list = sqrt(tpr * (1-fpr))  # getting geometric mean for tpr and fpr\n",
    "max_index = argmax(geomean_list) # getting the maximum mean index from the list of tpr and fpr coordinates\n",
    "\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "def plot_roc_curve(fpr, tpr, label=None): \n",
    "    plt.scatter(fpr[max_index], tpr[max_index], marker='o', c='navy', label='Optimal Threshold')\n",
    "    # plotting the optimal threshold coordinate as a dot\n",
    "    \n",
    "    plt.plot(fpr, tpr, color='cyan', alpha=0.5, linewidth=8, label=label) \n",
    "    # alpha parameter added for opacity - to show the optimal threshold dot that was hidden beneath it\n",
    "    \n",
    "    plt.plot([0, 1], [0, 1], color='white', lw=2, linestyle='--')\n",
    "    ax1 = fig.add_subplot(111)\n",
    "    plt.axis([0, 1, 0, 1])\n",
    "    plt.title('ROC Curve (1 month, negative boosting, RF)')\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.legend()\n",
    "    plt.show()\n",
    "\n",
    "    \n",
    "plot_roc_curve(fpr, tpr)\n",
    "\n",
    "print(\"Optimal Threshold: \", thresholds[max_index])\n",
    "print(\"Coordinate : (\", fpr[max_index],\", \", tpr[max_index], \")\")\n",
    "print(\"AUC score : \", roc_auc_score(y_test ,y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9998857368997847"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimal Threshold:  -1\n"
     ]
    }
   ],
   "source": [
    "plt.style.use('ggplot')\n",
    "\n",
    "# Get the 2nd column of the matrix of predicted probabilities for each data point\n",
    "#    The 2nd column stores the probalities of the positive class\n",
    "\n",
    "precisions, recalls, thresholds = precision_recall_curve(y_test, y_pred)\n",
    "\n",
    "fig = plt.figure(figsize=(10, 6))\n",
    "def plot_precision_recall_vs_threshold(precisions, recalls, thresholds): \n",
    "    plt.plot(thresholds, precisions[:-1], \"b--\",  linewidth=8, label=\"Precision\") \n",
    "    plt.plot(thresholds, recalls[:-1], \"g-\",  linewidth=3, label=\"Recall\") \n",
    "    plt.xlabel(\"Threshold\")\n",
    "    plt.legend(loc=\"lower right\")\n",
    "    plt.title('Precision-Recall Curve (1 month, negative boosting, RF)')\n",
    "    plt.ylim([0, 1.1])\n",
    "\n",
    "plot_precision_recall_vs_threshold(precisions, recalls, thresholds)\n",
    "plt.show()\n",
    "\n",
    "\n",
    "threshold_optimal = -1\n",
    "for i in range(len(precisions)):\n",
    "    if(precisions[i] == recalls[i]):\n",
    "        threshold_optimal = thresholds[i]\n",
    "\n",
    "print(\"Optimal Threshold: \", threshold_optimal)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Confusion Matrix:\n",
      "[[7864    5]\n",
      " [   8 1613]]\n"
     ]
    }
   ],
   "source": [
    "forest.fit(X_train, y_train)\n",
    "y_pred = forest.predict(X_test)\n",
    "\n",
    "print(\"Test Confusion Matrix:\")\n",
    "print(confusion_matrix(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Precision = 0.996910\n"
     ]
    }
   ],
   "source": [
    "precision_test = precision_score(y_test, y_pred) \n",
    "print(\"Test Precision = %f\" % precision_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Recall = 0.995065\n"
     ]
    }
   ],
   "source": [
    "recall_test = recall_score(y_test, y_pred)\n",
    "print(\"Test Recall = %f\" % recall_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test F1 Score = 0.995986\n"
     ]
    }
   ],
   "source": [
    "f1_test = f1_score(y_test, y_pred)\n",
    "print(\"Test F1 Score = %f\" % f1_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "TPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Getting_Twitter_data.ipynb",
   "provenance": []
  },
  "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}