random_forest_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Importing the required libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import warnings\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.exceptions import UndefinedMetricWarning\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.metrics import (accuracy_score, roc_curve,\n",
    "                             roc_auc_score, confusion_matrix,\n",
    "                             f1_score, log_loss, precision_recall_curve,\n",
    "                             average_precision_score, confusion_matrix, \n",
    "                             classification_report)\n",
    "\n",
    "warnings.filterwarnings('ignore', category=UndefinedMetricWarning)\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Loading the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_data = pd.read_csv('train_data.csv')\n",
    "test_data = pd.read_csv('test_data.csv')\n",
    "X_train, y_train = (train_data.drop('class of diagnosis', axis=1),\n",
    "                    train_data['class of diagnosis'])\n",
    "X_test, y_test = test_data.drop('class of diagnosis', axis=1), test_data['class of diagnosis']\n",
    "n_samples, n_features = X_train.shape\n",
    "seed = 123"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hyperparameter tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = RandomForestClassifier(random_state=seed, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters chosen:\n",
      "max_depth: 10\n",
      "max_features: 33\n",
      "min_samples_leaf: 1\n",
      "min_samples_split: 2\n",
      "n_estimators: 100\n"
     ]
    }
   ],
   "source": [
    "# Caution: computationaly intensive\n",
    "param_grid = {\n",
    "    'n_estimators': [10, 100, 500, 1000],\n",
    "    'max_depth': [3, 10, None],\n",
    "    'max_features': range(1, n_features + 1),\n",
    "    'min_samples_split': [2, 6, 10],\n",
    "    'min_samples_leaf': [1, 3, 5]\n",
    "}\n",
    "\n",
    "grid_search = GridSearchCV(model, param_grid=param_grid, n_jobs=-1, cv=5, scoring='f1')\n",
    "grid_search.fit(X_train, y_train)\n",
    "grid_model = grid_search.best_estimator_\n",
    "best = grid_search.best_params_\n",
    "print('Parameters chosen:')\n",
    "print(*['{}: {}'.format(parameter, value)\n",
    "        for parameter, value in zip(best.keys(), best.values())], sep='\\n')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Various metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluataion with 5-fold cross validation:\n",
      "accuracy  \t 0.7719115512678221\n",
      "neg_log_loss\t -0.5718360350717407\n",
      "f1        \t 0.5115011655011654\n",
      "roc_auc   \t 0.6811782212885155\n"
     ]
    }
   ],
   "source": [
    "print('Evaluataion with 5-fold cross validation:')\n",
    "for metric in ('accuracy', 'neg_log_loss', 'f1', 'roc_auc'):\n",
    "    score = np.mean(cross_val_score(grid_model,\n",
    "                                    X_train,\n",
    "                                    y_train,\n",
    "                                    scoring=metric,\n",
    "                                    cv=5,\n",
    "                                    n_jobs=-1))\n",
    "    print('{:<10}\\t {}'.format(metric, score))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=10, max_features=33, max_leaf_nodes=None,\n",
       "            min_impurity_split=1e-07, min_samples_leaf=1,\n",
       "            min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "            n_estimators=100, n_jobs=-1, oob_score=False, random_state=123,\n",
       "            verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_model.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1QAAAOFCAYAAABtA6PXAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n16EOQkdACgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb23d963668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "f_imp = pd.Series(grid_model.feature_importances_, X_train.columns)\n",
    "f_imp = f_imp.sort_values(ascending=False)\n",
    "plt.figure(figsize=(12,16))\n",
    "f_imp_plot = sns.barplot(y=f_imp.index, x=f_imp.values)\n",
    "plt.ylabel('Feature Importance Score')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### We can see a steady decrease in feature importance and a sharp drop after 'city_3'. Let's trim some features away and train the model on that."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluataion with 5-fold cross validation:\n",
      "accuracy  \t 0.7487563675687983\n",
      "neg_log_loss\t -0.5919006928583529\n",
      "f1        \t 0.47789775441949356\n",
      "roc_auc   \t 0.6636484593837535\n"
     ]
    }
   ],
   "source": [
    "high_importance = f_imp[f_imp > f_imp['city_3']].index\n",
    "X_train_new = X_train[high_importance]\n",
    "_, n_samples_new = X_train_new.shape\n",
    "best['max_features'] = min(n_samples_new, best['max_features'])\n",
    "model_new = RandomForestClassifier(random_state=seed, n_jobs=-1, **best)\n",
    "print('Evaluataion with 5-fold cross validation:')\n",
    "for metric in ('accuracy', 'neg_log_loss', 'f1', 'roc_auc'):\n",
    "    score = np.mean(cross_val_score(model_new,\n",
    "                                    X_train_new,\n",
    "                                    y_train,\n",
    "                                    scoring=metric,\n",
    "                                    cv=5,\n",
    "                                    n_jobs=-1))\n",
    "    print('{:<10}\\t {}'.format(metric, score))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### I don't think the drop in performance is worth it. Let's use the original model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Boxplots of predicted class probabilities and a confusion matrix for the training set. Just to make sure everything is working correctly so far."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEGCAYAAABsLkJ6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+v84yf4taJibrwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb201ebf6a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pred_train = grid_model.predict(X_train)\n",
    "pred_prob_train = grid_model.predict_proba(X_train)\n",
    "sns.boxplot(y_train, pred_prob_train.T[0])\n",
    "plt.ylabel(\"Predicted probability\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>predicted_negative</th>\n",
       "      <th>predicted_positive</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>actual_negative</th>\n",
       "      <td>178</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>actual_positive</th>\n",
       "      <td>0</td>\n",
       "      <td>81</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 predicted_negative  predicted_positive\n",
       "actual_negative                 178                   0\n",
       "actual_positive                   0                  81"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(data=confusion_matrix(y_train, pred_train),\n",
    "             columns=('predicted_negative', 'predicted_positive'),\n",
    "             index=('actual_negative', 'actual_positive'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Nicely separated. Time for predictions."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Performance on the test set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Class probabilities again, this time for the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEGCAYAAABsLkJ6AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAGeBJREFUeJzt3X+cXXV95/HXZIYQZgwy6CSGpRTFPt6A7q7GYMAgMiko\n+CtrscGKlmhcsWtLkTWNxVbBCG43FbdUoPVHu0CRNaIsSAHrLxBYxCTYSKp8FBKBYogDGfmRDEnm\nx/5xziU3N5l7T358770z5/18PObBPed77jmfIXfu53y/5/ujY2xsDDMzK58prQ7AzMxawwnAzKyk\nnADMzErKCcDMrKScAMzMSqqr1QEUNTDwjLsrmZntob6+6R3jlbkGYGZWUk4AZmYl5QRgZlZSTgBm\nZiXlBGBmVlJOAGZmJeUEYGZWUk4AZmYlNWEGgplZWitWXMvKlfe2NIbNmzcD0NPT09I4AI47bi4L\nF57V6jCScg3AzNrGtm1b2bZta6vDKI2OibIgjKeCMJv8liw5F4Dlyy9rcSSTh6eCMDOzXTgBmJmV\nlBOAmVlJuRdQE7VDLwton54WZehlYdbOXAMoIfe0MDNwDaCpFi48qy3ueN3TwszANQAzs9JyAjAz\nKyknADOzknICMDMrqaQPgSUtA+YD04BzImJVVdnbgL8AhoHrIuLzKWMxM7OdJasBSOoH5kTEPOBs\n4NKqsk7gcuAtwEnAuyUdnioWMzPbVcomoH7gRoCIWAscJqk7L3sx8HREPBERI8BdwCkJYzEzsxop\nm4BmAWuqtgeAmcD6/PV0SS8HHiWrBTxZ72S9vd10dXUmCrVcOjuzvN/XN73FkZjtzJ/N5kqZALbV\nbHcAYwARMSppMXAVsAn4N6Du0NTBwS0pYiylkZFRAAYGnmlxJGY782dz/6uXTFMmgA3AjOo4gI2V\njYj4DvAdAElXAo8kjMXMzGqkfAZwK7AAQNJsYF1EDFUKJd0iqU/SwWQ9hf4lYSxmZlYjWQ0gIlZL\nWiPpPrKunoslLQKeiogbgC+RfelPAT4ZEc+misXMzHaVdBxARCwFllbtur+q7BvAN1Je38zMxueR\nwGZmJeUEYGZWUk4AZmYl5QRgZlZSTgBmZiXlBGBmVlJOAGZmJeUEYGZWUk4AZmYl5QRgZlZSTgBm\nZiXlBGBmVlJJJ4NrJ5dcciGDg5taHUZbqPx/WLLk3BZH0h56ew/lggsubHUYZk1XmgQwOLiJJ598\nko4DDmp1KC03llf8Nj3tVdbGtg81PshskipNAgDoOOAgXvDyt7c6DGsjzz54U6tDMGsZPwMwMysp\nJwAzs5JyAjAzKyknADOzkkr6EFjSMmA+MA04JyJWVZX9MfAeYARYDfxpRIyljMfMzHZIVgOQ1A/M\niYh5wNnApVVlBwN/BpyYlx8DHJ8qFjMz21XKJqB+4EaAiFgLHCapOy/blv8cLKkL6AGeTBiLmZnV\nSNkENAtYU7U9AMwE1kfEc5I+Bfwc2AxcHxE/r3ey3t5uuro69zqYzk4/7rDd6+ycQl/f9FaHYez4\nO/W/R3OkTADbarY7gDF4vgnoz8mafp4Gvi3p1RHx4/FONji4b6NWR0ZG9+n9NnmNjIwyMPBMq8Mw\ndvyd+t9j/6mXTFPeFm8AZlTHAWzMXx8DPBgRAxGxFbgbmJ0wFjMzq5EyAdwKLACQNBtYFxGViVce\nBo6WdGC+/SrgFwljMTOzGsmagCJitaQ1ku4DhoHFkhYBT0XEDZIuBe6UNAzcHRE/SBWLmZntKuk4\ngIhYCiyt2nV/VdmVwJUpr29mZuMrzWygmzdvZmz7c5790XYytn2IzZs9/tDKqeEzAEmHNCMQMzNr\nriI1gJ9J+i7w5Yj4fuqAUunp6WHrSIfXA7CdPPvgTfT0dDc+0GwSKpIAjgBOA94v6bPA9cA/RsSG\npJGZmVlSDZuAImJ7RHwzIt4LvAt4M/CQpGsl9SWP0MzMkmhYA8jn73knsAh4IfBF4HSyWsF1wCkJ\n4zMzs0SKNAGtA24GPhYRP6ra/zVJ70wTlpmZpVZkJPAVEfGB6i9/SZ8EiIgzk0VmZmZJjVsDyOfz\nnw+8J5+yuWIq2fz+FyWOzczMEqrXBPQA2ZTOkK3aVbEZ+INkEZmVzCWXXMjg4KZWh9EWKv8fliw5\nt8WRtIfe3kO54IILk51/3ASQd/P8iqS7I+LhZBGYldzg4Cae3PQEUw4qzcD8cY1OyUZlDw79psWR\ntN7o0HDya9RrAvpq3sZ/l6RdxspHxBFJIzMrkSkHddF7mv+kbIfB2x5Jfo16txyVOtiJyaMwM7Om\nq5cA3iKp3nv/YT/HYmZmTVQvAby+wXudAMzMJrB6D4Hf18xAzMysuRo+BJb0KPli7tX8ENjMbGLz\nQ2Azs5Kq1wS0MX/5BNnI31eQ1QTuB65JH5qZmaVUZC6g64G5ZF/8PwXeAHw1ZVBmZpZekaGHL4yI\n06u2r5B0Z5GTS1pGNp/QNOCciFiV7/8PwLVVh76MbLbRrxQL28zM9lWRGsDPJb2ksiFpJvCLRm/K\nJ5ObExHzyJqQLq2URcRjEXFyRJwM/C7wCODV2s3MmqheL6A7ydr8pwHrJD0AjALHAqsKnLsfuBEg\nItZKOkxSd0RsqTluEXBjRDy7F/GbmdleqtcE9Bd1yorUHGYBa6q2B4CZwPqa4z4InNroZL293XR1\ndRa47O51dhYJ2cqos3MKfX3TW3p9s91J/dms1wvojsprSS8ADs03DwSuAl7X4NzbarY7qBlPIGke\n8EhEPN0o0MHB2orDnhkZGWVs+xDPPuiWprGR7J+mo3NqiyNpvbHtQ4yMdDMw8EzLYhgZGW3Zta29\njYyM7vNns14CKbIm8J8BF5B98Q+RLQhTpBvoBmBGdRzAxppj3gx8s8C59llv76GNDyqJwcHnAOg9\nuLvFkbSDbn82rLSK9AJ6J9kX+bcioj9fB/jwAu+7FbgYuFLSbGBdRAzVHDMX+PqeBLy3Ui6qMNFU\nFttYvvyyFkdiZq1UpPFxS0RsI08WEXE9cHr9t0BErAbWSLoP+DvgfEmLJL2j6rBZwK/2PGwzM9tX\nRWoAGyV9APiFpGuAn1CsBkBELAWWVu26v6b8FUUDNTOz/atIDeBs4NvAeWT9/w8H3pUyKDMzS69h\nAoiI54Ae4E1kU0FcERH313+XmZm1u4YJQNJnyQZ0nQEsBG6RdEnqwMzMLK0izwD6gWMjYjuApAOB\ne8i6hpqZ2QRV5BnAr4CRqu3twC+TRGNmZk1Tby6gT+UvnwZWSrqDbC6gNwBrmxCbmZklVK8JqHLX\nH/lPxc3A3k/KY2ZmbaHeXEAXVV7ncwGJrAYQu5nR08zMJpgivYDOAB4ErgC+QDYgrOFIYDMza29F\negEtAf5TRPwanl/N63qyuX7MzGyCKtILaGvlyx+y1bzYdapnMzObYIrUAJ6WtAS4Ld8+DXgqXUhm\n5bJ582ZGtw4zeNsjrQ7F2sjo0DCbRzcnvUaRGsBi4Cjg6vznyHyfmZlNYHVrAJI6gNdGxIeaFI9Z\n6fT09LBtynZ6Tzui1aFYGxm87RF6DupJeo26NYCIGAM+IqlIU5GZmU0gRb7YnwB+li/ssrWyMyL+\nMFlUZmaWXJEEcEv+Y2Zmk0iR9QCuAlYCW4DNwA/zfWZmNoEVGQm8HPgm8HvsWA/g4tSBmZlZWkWa\ngE4Bjq5aD2AqcC/w8UZvlLQMmA9MA86JiFVVZYcD1+ZlP3ZPIzOz5ioyDuCJypd/bphsjYC6JPUD\ncyJiHtm6wpfWHHIx8MmImAuMSvrtgjGbmdl+UKQG8Jike4EfAB1k6wGsq6wXEBGfGOd9/WRLSRIR\nayUdJqm7aibR10TE2Xn5f9uXX8LMzPZckQTwCLC+avvmgueeBayp2h4AZgLrJR0CbJH0JeAY4AcR\n8ef1Ttbb201Xl5ch2B86O7OKX1/f9BZHYrDj38OsVmfnlKR/pw0TQJ07/EZqJ4zrAMby1wcCRwO/\nDzwG/LOkt0XEN8c72eCglyDYX0ZGRgEYGHimxZEY7Pj3MKs1MjK6z3+n9RJIyluPDcCM6jiAjfnr\nJ4D1EfFwRAwD3yarCZiZWZOkTAC3AgsAJM0G1kXEEEBEjAAPS3pZfuxcdl520szMEks2x09ErJa0\nJp9CYhhYLGkR8FRE3ACcD/y9pB6yReZvShWLmZntatwEIGmUHW32tbZHxLRGJ4+IpcDSql33V5U9\nCJxaME4zM9vP6tUADiB7cPtxst483yVrMnoT2QLxZmY2gY2bAPJ2eiT1R8RFVUUrJN02ztvMzGyC\nKPIMoFvSB4G78+3XsXPvHjMzm4CKJID3AJ8E/pisSejfgPemDMrMzNIrMhDs55LOBg6LCK9abWY2\nSRSZDvp04CHgW/n230j6vdSBmZlZWkUGgn0cmAM8nm9fBHwsWURmZtYURRLAcxExUNmIiE3AULqQ\nzMysGYo8BN4q6USgQ1IvcCbwXNqwzMwstSIJ4MPA54H/DDwI3AX815RBmZlZekUSwJER8dbqHZIW\nkK0TYGZmE1S9uYCOBI4C/lrS+WRjACCbIuJz5Kt9mZnZxFSvBjCLrL3/SKB6UZhR4AsJYzIzsyao\nNxfQPcA9km4BboyIMQBJXfkiLmZmNoEV6QZ6EDvP1X+3pIWJ4jEzsyYpkgDOI5sPqOKNwEfThGNm\nZs1SdCDYU5WN/HXtgu9mZjbBFOkGukbSCuAOsoTxRmB10qjMzCy5IgngT4HfB07It68BvpYsIjMz\na4p64wBmRcQG4KXAqvyn4qXAukYnl7QMmA9MA86JiFVVZb8EHgVG8l1nRcRjexi/mZntpXo1gM8C\n7yZbC7jWGPCyeieW1A/MiYh5kl4JXAGcVHPY6RHx7B7Ea2Zm+0m9cQDvzv/70r08dz/5aOGIWCvp\nMEndEbFlL89nZmb7Ub0moH+o874pEbGowblnAWuqtgeAmcD6qn1flPRbZOsNf6wy2Gx3enu76erq\nbHBJK6KzM+v81dc3vcWRGOz49zCr1dk5Jenfab0moMoi8CcALwK+R9YL6FR2/hIfT21X0Q6ypqOK\nTwDfBn4NfJ3sQfOK8U42ODjxKw4rVlzLypX3tjoMBgc3AbBo0ftaGsdxx81l4cKzWhpDOxgZGW11\nCNamRkZGGRh4Zp/OUS+B1GsC+jKApHdGxOmV/ZIuA/5vgetuAGZUxwFsrDr/1VXn/BbwigLntP1g\n6tQDWx2CmbWBIt1AD5f0wqrBYAfT4AFw7lbgYuBKSbOBdRExBCBpOnAz2UPgLcCJFEsqE9rChWe1\nxR3vAw/8FICjjz62xZGYWSsVSQBXAg9JWk/WjHMk8OlGb4qI1ZLWSLoPGAYWS1oEPBURN+SDy+6S\ntAX4MXD9Xv4OtoduvPHrgBOAWdk1TAARcYWkfwJeTpYAHoqI3xQ5eUQsBZZW7bq/quxy4PI9C9f2\n1QMP/JSInz3/2kmgPYwODTN4m9dYGt2WDQuaMtUdPkaHhrOpOBNqmAAkvZjsge3MiDhT0jsk3RMR\nj6cNzVKo3P1XXjsBtF5v76GtDqFtDD6XdVDoPeiQFkfSBg5K/9ko0gT0JeBfgNlV+64C3pQkIrOS\nueCCC1sdQttYsuRcAJYvv6zFkZRDkQ7I0yPiCmA7QETcALgbyQS1YMEZu31tZuVTpAYwVdIB5H34\nJc0AupNGZckcffSxSMc8/9rMyqtIArgCWAnMknQT8FqyGUJtgvKdv5lBsV5A10m6AziOrBZwTj5L\nqJmZTWB1E4CkDuD6iDiDfGI3m/g8DsDMoEECiIgxSb+QtBj4f8DWqrKG6wFY+/E4ADOrKPIM4Mzd\n7Gu4HoC1J48DMLOKIs8A9nY9ADMza2PjjgOQNF3Sckk3SVoqqUhtwdqcxwGYWUW9gWBfIFuv94uA\ngE81JSJLqjIOQDrGzT9mJVfvrv63I+IPACTdQrYgjE0CvvM3M6ifAKp7/IxI8rJFk4Tv/M0M6jcB\n1a7PO+56vWZmNvHUqwG8TlL1BOUz8u0OYCwijkgbmpmZpVQvAahpUZiZWdPVWxT+4WYGYmZmzVVk\nPQAzM5uEkg7ukrQMmA9MI5tFdNVujvkMcEJEnJwyFjMz21myGoCkfmBORMwDzgYu3c0xxwInpYrB\nzMzGl7IJqJ98CumIWAscJql2JbG/Bi5IGIOZmY0jZRPQLGBN1fYAMBNYDyBpEfB9oNDD5t7ebrq6\nOvdziGbWTjo7s3vSvr7pLY6kHFImgG012x3sWFf4UOA9wGnA4UVONji4Zb8GZ2btZ2Qkm3BgYOCZ\nFkcyedRLpimbgDYAM6rjADbmr+eT1RDuAm4AZkv6XMJYzMysRsoEcCuwAEDSbGBdRAwBRMT1EfGK\niDgeeAdwX0R8JGEsZmZWI1kCiIjVwBpJ9wF/B5wvaZGkd6S6ppmZFZd0HEBELAWWVu26fzfH/BI4\nOWUcZma2K48ENjMrKScAM7OScgIwMyspJwAzs5JyAjAzKyknADOzknICMDMrKScAM7OScgIwMysp\nJwAzs5JyAjAzKyknADOzknICMDMrKScAM7OScgIwMyspJwAzs5JyAjAzKyknADOzknICMDMrKScA\nM7OSSroovKRlwHxgGnBORKyqKvsg8H5gDPgJ8KGIGEsZj5mZ7ZCsBiCpH5gTEfOAs4FLq8q6gXcB\nr4+IE4DfAU5IFYuZme0qZRNQP3AjQESsBQ7Lv/iJiC0RMT8ituf7pgOPJ4zFzMxqpGwCmgWsqdoe\nAGYC6ys7JH0M+AiwPCLW1TtZb283XV2dKeI0szbR2Zndk/b1TW9xJOWQMgFsq9nuIGvvf15E/A9J\nlwH/LOlHEfGD8U42OLglQYhm1k5GRkYBGBh4psWRTB71kmnKJqANwIzqOICNAJIOlXQyZM1BwC3A\n8QljMTOzGikTwK3AAgBJs4F1ETFUdd0vS+rJt+cCkTAWMzOrkawJKCJWS1oj6T5gGFgsaRHwVETc\nIOki4PuShsmeFdyUKhYzM9tV0nEAEbEUWFq16/6qsquBq1Ne38zMxueRwGZmJeUEYGZWUk4AZmYl\n5QRgZlZSTgBmZiXlBGBmVlIdY2MTYwbmgYFnJkagZhPUihXXsnLlvS2NYXBwEwC9vYe2NA6A446b\ny8KFZ7U6jH3W1ze9Y7yypOMAzMz2xNSpB7Y6hFJxDcDMbBKrVwPwMwAzs5JyAjAzKyknADOzknIC\nMDMrKScAM7OScgIwMyspJwAzs5JyAjAzK6kJMxDMzMz2L9cAzMxKygnAzKyknADMzErKCcDMrKSc\nAMzMSsoJwMyspJwAzMxKyiuCTXKSlgHzgWnAORGxqqrsBOCzedk3IuLTrYnSykrSK4Ebgc9FxOdr\nyvz5TMw1gElMUj8wJyLmAWcDl9YcchVwJjAHeJuko5ocopWYpB7gb4HvjnOIP5+JOQFMbv1kd1dE\nxFrgMEndAJJeBmyKiEcjYhS4GXhjyyK1MtoKvBn4VW2BP5/N4QQwuc0CBqq2B4CZ45T9GnhJk+Iy\nIyKGI2JonGJ/PpvACWBy21az3QGMFSgzazV/PpvACWBy2wDMqNruAzaOU/YSdlMVN2sRfz6bwAlg\ncrsVWAAgaTawrlLljoh/Bw6QdISkTuCt+fFmLefPZ3N4OuhJTtJfAacCw8Bi4DXAUxFxg6STgL8h\nq1r/U0TU9hIyS0bSa8i6eR4JbAceA24C1vvz2RxOAGZmJeUmIDOzknICMDMrKScAM7OScgIwMysp\nJwAzs5JyArAJRdLtkk5p8jXPlfSQpLfUOeYUSbfnr/9X3sWxJSS9StLftur6NnF4OmizxhYAH46I\n24ocHBHnJY6n0fX/FfiTVsZgE4MTgLUtSR8nmy1yKnBd9UAgSR3A5cCrgE7g3og4V9ILgK8AvWSf\n7+9ExF9KOhP4KPBsvv8DERE113s/8EfAENnkY4uB95INnvsrSVMj4qaq4/8LcDHZFAVRtf924NNk\n0xzvEmN+zCVko1sfBh4nm/rgImATsCwvmwm8KyLWSJpLNp33cH6ZD0fEWknnAn+Y/17b8niPAT4d\nESfurjwiKtOBWMm5CcjakqTjgbcDJwGvA06W1Ft1yCHAQ8A84HjgTfniIqcCUyPi9fn7NkqaClwA\n/ElE9JPdHR9ec70jgE8BvxsRJwGPAOfni5T8K/Dfq7/8c58HzoiIU4GR3fwau41Rksi+qOcCC/Pf\ncTQiRoCDgZ9FxMnAdcD783NdncfzBrLRs5fn+5cBb82Pv6T29ypQbiXmBGDt6njgrogYiYjtEfH2\niBisKn8W+C3gTuD7ZJOFvRj4IXCUpK8B7wOuiYhtwP8BrpF0MXBgRNQuQjIbWBkRT+fb3yVbiGS3\nJL0IOCgiHqg6vtZ4Mb4yv9ZQPjfT7TXv+17+30eAQyUdAsyIiHurrlV5xnAd8C1JfwkMRMTqmnM1\nKrcScwKwdlbv8/lu4NXA/Pzudj1ARGwgawK5HPiPwE8kHRIRnyFbIGc98AVJH21w7Q5gtEF5o1h3\nG2N+bPW5a6+zvc51KvvGACLiQ8AZwFPATZIWVh/YqNzKzQnA2tUPgX5JB+Q/d0iaVVX+IuCnEbEt\nby56KXCgpDcCp0XE7RHxEbJ276MkfYbsDvhLwHLgxJrrrQLmSDo43z4tj2E8TwLb8+Yc2P1qVbuN\nkex5wavz36sHqNurKSJ+Azwu6bXVsUnqlfRJ4KGIuAz4R7JmLwAalZv5IbC1pYj4Yd6McxfZjcpX\nImLDju9bvgrcLOl7wI+A/0n2kPQ04H9LWkL28PgW4D7g9cBdkp7N959Xc71/l/QJ4DuStgKPAkvr\nxDcm6TzgBknrgAfY9YZqvBhPJJvaeHX+vjvZ8XB3PGcDn5M0nB/7RxExmC/xea+kzflxi4Ej8hjH\nKzcDPBuoWdNJ6iL7Qr86IrZLujl/vaLFoVnJuAnIrMkiYpisOehHku4BngC+3tqorIxcAzAzKynX\nAMzMSsoJwMyspJwAzMxKygnAzKyknADMzErq/wN67tG3jjiLZAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb201731278>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pred = grid_model.predict(X_test)\n",
    "pred_prob = grid_model.predict_proba(X_test)\n",
    "sns.boxplot(y_test, pred_prob.T[0])\n",
    "plt.ylabel(\"Predicted probability\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Precision, recall, f1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.82      0.94      0.88        50\n",
      "        1.0       0.62      0.33      0.43        15\n",
      "\n",
      "avg / total       0.78      0.80      0.78        65\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>predicted_negative</th>\n",
       "      <th>predicted_positive</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>actual_negative</th>\n",
       "      <td>47</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>actual_positive</th>\n",
       "      <td>10</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 predicted_negative  predicted_positive\n",
       "actual_negative                  47                   3\n",
       "actual_positive                  10                   5"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(data=confusion_matrix(y_test, pred),\n",
    "             columns=('predicted_negative', 'predicted_positive'),\n",
    "             index=('actual_negative', 'actual_positive'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Confusion matrices and precision/recall/f1 for different probability cuttofs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Cutoff at 0.55"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.84      0.94      0.89        50\n",
      "        1.0       0.67      0.40      0.50        15\n",
      "\n",
      "avg / total       0.80      0.82      0.80        65\n",
      "\n"
     ]
    }
   ],
   "source": [
    "pred_55 = [0 if prob > 0.55 else 1 for prob in pred_prob.T[0]]\n",
    "print(classification_report(y_test, pred_55))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>predicted_negative</th>\n",
       "      <th>predicted_positive</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>actual_negative</th>\n",
       "      <td>47</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>actual_positive</th>\n",
       "      <td>9</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 predicted_negative  predicted_positive\n",
       "actual_negative                  47                   3\n",
       "actual_positive                   9                   6"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(data=confusion_matrix(y_test, pred_55),\n",
    "             columns=('predicted_negative', 'predicted_positive'),\n",
    "             index=('actual_negative', 'actual_positive'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Cutoff at 0.65"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.88      0.84      0.86        50\n",
      "        1.0       0.53      0.60      0.56        15\n",
      "\n",
      "avg / total       0.80      0.78      0.79        65\n",
      "\n"
     ]
    }
   ],
   "source": [
    "pred_65 = [0 if prob > 0.65 else 1 for prob in pred_prob.T[0]]\n",
    "print(classification_report(y_test, pred_65))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>predicted_negative</th>\n",
       "      <th>predicted_positive</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>actual_negative</th>\n",
       "      <td>42</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>actual_positive</th>\n",
       "      <td>6</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 predicted_negative  predicted_positive\n",
       "actual_negative                  42                   8\n",
       "actual_positive                   6                   9"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.DataFrame(data=confusion_matrix(y_test, pred_65),\n",
    "             columns=('predicted_negative', 'predicted_positive'),\n",
    "             index=('actual_negative', 'actual_positive'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ROC curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb201736f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Area under curve: 0.716\n"
     ]
    }
   ],
   "source": [
    "fpr, tpr, thresholds = roc_curve(y_test, pred_prob.T[1])\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(fpr, tpr, 'b')\n",
    "plt.plot([0,1],[0,1], color='red', linestyle='--')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.show()\n",
    "roc_auc = roc_auc_score(y_test, pred_prob.T[1])\n",
    "print(''.join(('Area under curve: ', str(roc_auc))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Precision-recall curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEVCAYAAADpbDJPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nlABERGLq/wMJ1NniVcfJLQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb20160f7f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Area under curve: 0.554817265957\n"
     ]
    }
   ],
   "source": [
    "precision, recall, thresholds = precision_recall_curve(y_test, pred_prob.T[1])\n",
    "plt.title('Precision-Recall Curve')\n",
    "plt.plot(recall, precision, 'b')\n",
    "plt.ylabel('Precision')\n",
    "plt.xlabel('Recall')\n",
    "plt.show()\n",
    "prec_recall = average_precision_score(y_test, pred_prob.T[1])\n",
    "print(''.join(('Area under curve: ', str(prec_recall))))"
   ]
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}