eye-blink-detection-3-ml-model-part2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "# EYE BLINK DETECTION:\n",
    "\n",
    "# 3) Machine Learning Model - part2\n",
    "\n",
    "For this part we will implement the ML model described in [the paper by Soukupova and Cech (2016)](http://http://vision.fe.uni-lj.si/cvww2016/proceedings/papers/05.pdf).\n",
    "\n",
    "We will use the output dataframes of part-1. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import sklearn methods \n",
    "from sklearn.metrics import accuracy_score, roc_curve, confusion_matrix, classification_report, roc_auc_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.svm import LinearSVC,SVC\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Read training set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>subject</th>\n",
       "      <th>external_factors</th>\n",
       "      <th>facial_actions</th>\n",
       "      <th>frame_no</th>\n",
       "      <th>face_detected</th>\n",
       "      <th>left_ear</th>\n",
       "      <th>right_ear</th>\n",
       "      <th>avg_ear</th>\n",
       "      <th>mar</th>\n",
       "      <th>moe</th>\n",
       "      <th>...</th>\n",
       "      <th>right_sop</th>\n",
       "      <th>avg_sop</th>\n",
       "      <th>closeness</th>\n",
       "      <th>blink_no</th>\n",
       "      <th>blink_start_frame</th>\n",
       "      <th>blink_end_frame</th>\n",
       "      <th>reserved_for_calibration</th>\n",
       "      <th>perclos</th>\n",
       "      <th>closeness_annot</th>\n",
       "      <th>blink_annot</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>26122013_223310_cam</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.258551</td>\n",
       "      <td>0.251491</td>\n",
       "      <td>0.255021</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.421133</td>\n",
       "      <td>0.417291</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>26122013_223310_cam</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0.269800</td>\n",
       "      <td>0.239934</td>\n",
       "      <td>0.254867</td>\n",
       "      <td>0.032141</td>\n",
       "      <td>0.126110</td>\n",
       "      <td>...</td>\n",
       "      <td>0.468100</td>\n",
       "      <td>0.432625</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>26122013_223310_cam</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0.230599</td>\n",
       "      <td>0.293610</td>\n",
       "      <td>0.262104</td>\n",
       "      <td>0.022217</td>\n",
       "      <td>0.084763</td>\n",
       "      <td>...</td>\n",
       "      <td>0.415718</td>\n",
       "      <td>0.428555</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>26122013_223310_cam</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0.269800</td>\n",
       "      <td>0.293610</td>\n",
       "      <td>0.281705</td>\n",
       "      <td>0.023256</td>\n",
       "      <td>0.082554</td>\n",
       "      <td>...</td>\n",
       "      <td>0.468100</td>\n",
       "      <td>0.454746</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>None</td>\n",
       "      <td>26122013_223310_cam</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>0.258551</td>\n",
       "      <td>0.281942</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.450859</td>\n",
       "      <td>0.437356</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 27 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  subject external_factors       facial_actions  frame_no  face_detected  \\\n",
       "0       1             None  26122013_223310_cam         1              1   \n",
       "1       1             None  26122013_223310_cam         2              1   \n",
       "2       1             None  26122013_223310_cam         3              1   \n",
       "3       1             None  26122013_223310_cam         4              1   \n",
       "4       1             None  26122013_223310_cam         5              1   \n",
       "\n",
       "   left_ear  right_ear   avg_ear       mar       moe  ...  right_sop  \\\n",
       "0  0.258551   0.251491  0.255021  0.000000  0.000000  ...   0.421133   \n",
       "1  0.269800   0.239934  0.254867  0.032141  0.126110  ...   0.468100   \n",
       "2  0.230599   0.293610  0.262104  0.022217  0.084763  ...   0.415718   \n",
       "3  0.269800   0.293610  0.281705  0.023256  0.082554  ...   0.468100   \n",
       "4  0.258551   0.281942  0.270246  0.000000  0.000000  ...   0.450859   \n",
       "\n",
       "    avg_sop  closeness  blink_no  blink_start_frame  blink_end_frame  \\\n",
       "0  0.417291          0         0                  1                1   \n",
       "1  0.432625          0         0                  1                1   \n",
       "2  0.428555          0         0                  1                1   \n",
       "3  0.454746          0         0                  1                1   \n",
       "4  0.437356          0         0                  1                1   \n",
       "\n",
       "   reserved_for_calibration  perclos  closeness_annot  blink_annot  \n",
       "0                      True     -1.0                0            0  \n",
       "1                      True     -1.0                0            0  \n",
       "2                      True     -1.0                0            0  \n",
       "3                      True     -1.0                0            0  \n",
       "4                      True     -1.0                0            0  \n",
       "\n",
       "[5 rows x 27 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# read dataset\n",
    "df = pd.read_pickle('../input/eye-blink-detection-3-ml-model-part1/train/training_set.pkl')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "construct (n x 13) matrix of training set. Columns are defining 13 frame window (6 previous frames + current frame + 6 next frames)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:15: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  from ipykernel import kernelapp as app\n"
     ]
    },
    {
     "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>subject</th>\n",
       "      <th>blink_annot</th>\n",
       "      <th>ear-6</th>\n",
       "      <th>ear-5</th>\n",
       "      <th>ear-4</th>\n",
       "      <th>ear-3</th>\n",
       "      <th>ear-2</th>\n",
       "      <th>ear-1</th>\n",
       "      <th>ear0</th>\n",
       "      <th>ear1</th>\n",
       "      <th>ear2</th>\n",
       "      <th>ear3</th>\n",
       "      <th>ear4</th>\n",
       "      <th>ear5</th>\n",
       "      <th>ear6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.255021</td>\n",
       "      <td>0.254867</td>\n",
       "      <td>0.262104</td>\n",
       "      <td>0.281705</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.294951</td>\n",
       "      <td>0.304514</td>\n",
       "      <td>0.295574</td>\n",
       "      <td>0.294976</td>\n",
       "      <td>0.298302</td>\n",
       "      <td>0.300481</td>\n",
       "      <td>0.295888</td>\n",
       "      <td>0.294953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.254867</td>\n",
       "      <td>0.262104</td>\n",
       "      <td>0.281705</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.294951</td>\n",
       "      <td>0.304514</td>\n",
       "      <td>0.295574</td>\n",
       "      <td>0.294976</td>\n",
       "      <td>0.298302</td>\n",
       "      <td>0.300481</td>\n",
       "      <td>0.295888</td>\n",
       "      <td>0.294953</td>\n",
       "      <td>0.256632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.262104</td>\n",
       "      <td>0.281705</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.294951</td>\n",
       "      <td>0.304514</td>\n",
       "      <td>0.295574</td>\n",
       "      <td>0.294976</td>\n",
       "      <td>0.298302</td>\n",
       "      <td>0.300481</td>\n",
       "      <td>0.295888</td>\n",
       "      <td>0.294953</td>\n",
       "      <td>0.256632</td>\n",
       "      <td>0.256280</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.281705</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.294951</td>\n",
       "      <td>0.304514</td>\n",
       "      <td>0.295574</td>\n",
       "      <td>0.294976</td>\n",
       "      <td>0.298302</td>\n",
       "      <td>0.300481</td>\n",
       "      <td>0.295888</td>\n",
       "      <td>0.294953</td>\n",
       "      <td>0.256632</td>\n",
       "      <td>0.256280</td>\n",
       "      <td>0.267013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.270246</td>\n",
       "      <td>0.294951</td>\n",
       "      <td>0.304514</td>\n",
       "      <td>0.295574</td>\n",
       "      <td>0.294976</td>\n",
       "      <td>0.298302</td>\n",
       "      <td>0.300481</td>\n",
       "      <td>0.295888</td>\n",
       "      <td>0.294953</td>\n",
       "      <td>0.256632</td>\n",
       "      <td>0.256280</td>\n",
       "      <td>0.267013</td>\n",
       "      <td>0.266355</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",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71255</th>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0.253677</td>\n",
       "      <td>0.247394</td>\n",
       "      <td>0.251768</td>\n",
       "      <td>0.227690</td>\n",
       "      <td>0.245400</td>\n",
       "      <td>0.271400</td>\n",
       "      <td>0.229162</td>\n",
       "      <td>0.249375</td>\n",
       "      <td>0.242212</td>\n",
       "      <td>0.236984</td>\n",
       "      <td>0.255580</td>\n",
       "      <td>0.231446</td>\n",
       "      <td>0.249920</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71256</th>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0.247394</td>\n",
       "      <td>0.251768</td>\n",
       "      <td>0.227690</td>\n",
       "      <td>0.245400</td>\n",
       "      <td>0.271400</td>\n",
       "      <td>0.229162</td>\n",
       "      <td>0.249375</td>\n",
       "      <td>0.242212</td>\n",
       "      <td>0.236984</td>\n",
       "      <td>0.255580</td>\n",
       "      <td>0.231446</td>\n",
       "      <td>0.249920</td>\n",
       "      <td>0.227307</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71257</th>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0.251768</td>\n",
       "      <td>0.227690</td>\n",
       "      <td>0.245400</td>\n",
       "      <td>0.271400</td>\n",
       "      <td>0.229162</td>\n",
       "      <td>0.249375</td>\n",
       "      <td>0.242212</td>\n",
       "      <td>0.236984</td>\n",
       "      <td>0.255580</td>\n",
       "      <td>0.231446</td>\n",
       "      <td>0.249920</td>\n",
       "      <td>0.227307</td>\n",
       "      <td>0.254391</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71258</th>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0.227690</td>\n",
       "      <td>0.245400</td>\n",
       "      <td>0.271400</td>\n",
       "      <td>0.229162</td>\n",
       "      <td>0.249375</td>\n",
       "      <td>0.242212</td>\n",
       "      <td>0.236984</td>\n",
       "      <td>0.255580</td>\n",
       "      <td>0.231446</td>\n",
       "      <td>0.249920</td>\n",
       "      <td>0.227307</td>\n",
       "      <td>0.254391</td>\n",
       "      <td>0.262248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>71259</th>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0.245400</td>\n",
       "      <td>0.271400</td>\n",
       "      <td>0.229162</td>\n",
       "      <td>0.249375</td>\n",
       "      <td>0.242212</td>\n",
       "      <td>0.236984</td>\n",
       "      <td>0.255580</td>\n",
       "      <td>0.231446</td>\n",
       "      <td>0.249920</td>\n",
       "      <td>0.227307</td>\n",
       "      <td>0.254391</td>\n",
       "      <td>0.262248</td>\n",
       "      <td>0.246758</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>71260 rows × 15 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      subject  blink_annot     ear-6     ear-5     ear-4     ear-3     ear-2  \\\n",
       "0           1            0  0.255021  0.254867  0.262104  0.281705  0.270246   \n",
       "1           1            0  0.254867  0.262104  0.281705  0.270246  0.294951   \n",
       "2           1            0  0.262104  0.281705  0.270246  0.294951  0.304514   \n",
       "3           1            0  0.281705  0.270246  0.294951  0.304514  0.295574   \n",
       "4           1            0  0.270246  0.294951  0.304514  0.295574  0.294976   \n",
       "...       ...          ...       ...       ...       ...       ...       ...   \n",
       "71255      11            0  0.253677  0.247394  0.251768  0.227690  0.245400   \n",
       "71256      11            0  0.247394  0.251768  0.227690  0.245400  0.271400   \n",
       "71257      11            0  0.251768  0.227690  0.245400  0.271400  0.229162   \n",
       "71258      11            0  0.227690  0.245400  0.271400  0.229162  0.249375   \n",
       "71259      11            0  0.245400  0.271400  0.229162  0.249375  0.242212   \n",
       "\n",
       "          ear-1      ear0      ear1      ear2      ear3      ear4      ear5  \\\n",
       "0      0.294951  0.304514  0.295574  0.294976  0.298302  0.300481  0.295888   \n",
       "1      0.304514  0.295574  0.294976  0.298302  0.300481  0.295888  0.294953   \n",
       "2      0.295574  0.294976  0.298302  0.300481  0.295888  0.294953  0.256632   \n",
       "3      0.294976  0.298302  0.300481  0.295888  0.294953  0.256632  0.256280   \n",
       "4      0.298302  0.300481  0.295888  0.294953  0.256632  0.256280  0.267013   \n",
       "...         ...       ...       ...       ...       ...       ...       ...   \n",
       "71255  0.271400  0.229162  0.249375  0.242212  0.236984  0.255580  0.231446   \n",
       "71256  0.229162  0.249375  0.242212  0.236984  0.255580  0.231446  0.249920   \n",
       "71257  0.249375  0.242212  0.236984  0.255580  0.231446  0.249920  0.227307   \n",
       "71258  0.242212  0.236984  0.255580  0.231446  0.249920  0.227307  0.254391   \n",
       "71259  0.236984  0.255580  0.231446  0.249920  0.227307  0.254391  0.262248   \n",
       "\n",
       "           ear6  \n",
       "0      0.294953  \n",
       "1      0.256632  \n",
       "2      0.256280  \n",
       "3      0.267013  \n",
       "4      0.266355  \n",
       "...         ...  \n",
       "71255  0.249920  \n",
       "71256  0.227307  \n",
       "71257  0.254391  \n",
       "71258  0.262248  \n",
       "71259  0.246758  \n",
       "\n",
       "[71260 rows x 15 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# only get necessary columns\n",
    "ear = df[['subject','frame_no','avg_ear','blink_annot']]\n",
    "\n",
    "# group by subject\n",
    "user_list = list(ear['subject'].unique())\n",
    "list_of_dfs=[]\n",
    "for user in user_list:\n",
    "    list_of_dfs.append(ear.groupby('subject').get_group(user))\n",
    "\n",
    "\n",
    "# construct a df with columns depicting 13 frame window (6 previous frames + current frame + 6 next frames)\n",
    "list_of_13dfs = []\n",
    "for subject_df in list_of_dfs:\n",
    "    for i in range(-6,7):\n",
    "        subject_df['ear'+str(i)] = subject_df.shift(periods=i*(-1))['avg_ear']\n",
    "    subject_df=subject_df[6:-6]\n",
    "    list_of_13dfs.append(subject_df)\n",
    "    \n",
    "# concat results\n",
    "ear13_df = pd.concat(list_of_13dfs)\n",
    "ear13_df.drop(columns=['frame_no', 'avg_ear'],inplace=True)\n",
    "ear13_df.reset_index(drop=True, inplace=True)\n",
    "ear13_df"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Some analysis on the dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 71260 entries, 0 to 71259\n",
      "Data columns (total 15 columns):\n",
      " #   Column       Non-Null Count  Dtype  \n",
      "---  ------       --------------  -----  \n",
      " 0   subject      71260 non-null  object \n",
      " 1   blink_annot  71260 non-null  int64  \n",
      " 2   ear-6        71260 non-null  float64\n",
      " 3   ear-5        71260 non-null  float64\n",
      " 4   ear-4        71260 non-null  float64\n",
      " 5   ear-3        71260 non-null  float64\n",
      " 6   ear-2        71260 non-null  float64\n",
      " 7   ear-1        71260 non-null  float64\n",
      " 8   ear0         71260 non-null  float64\n",
      " 9   ear1         71260 non-null  float64\n",
      " 10  ear2         71260 non-null  float64\n",
      " 11  ear3         71260 non-null  float64\n",
      " 12  ear4         71260 non-null  float64\n",
      " 13  ear5         71260 non-null  float64\n",
      " 14  ear6         71260 non-null  float64\n",
      "dtypes: float64(13), int64(1), object(1)\n",
      "memory usage: 8.2+ MB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# df info\n",
    "print(ear13_df.info())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TOTAL COUNT: 71260\n",
      "POSITIVE COUNT: 3709\n",
      "NEGATIVE COUNT: 67551\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 216x216 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# number of classes\n",
    "plt.figure(figsize =(3,3))\n",
    "plt.hist(ear13_df.blink_annot)\n",
    "total_count = len(ear13_df)\n",
    "pos_count = ear13_df.loc[ear13_df['blink_annot'] == 1, 'blink_annot'].count()\n",
    "neg_count = ear13_df.loc[ear13_df['blink_annot'] == 0, 'blink_annot'].count()\n",
    "print(\"TOTAL COUNT: {}\\nPOSITIVE COUNT: {}\\nNEGATIVE COUNT: {}\\n\".format(total_count,pos_count,neg_count))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "subject        0\n",
      "blink_annot    0\n",
      "ear-6          0\n",
      "ear-5          0\n",
      "ear-4          0\n",
      "ear-3          0\n",
      "ear-2          0\n",
      "ear-1          0\n",
      "ear0           0\n",
      "ear1           0\n",
      "ear2           0\n",
      "ear3           0\n",
      "ear4           0\n",
      "ear5           0\n",
      "ear6           0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "# Look for missing values\n",
    "print(ear13_df.isnull().sum())  "
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Now implement SVM:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/joblib/externals/loky/process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The best parameters are {'C': 10, 'class_weight': None, 'gamma': 'scale', 'kernel': 'rbf', 'max_iter': 5000} with a score of 0.9779\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/sklearn/svm/_base.py:231: ConvergenceWarning: Solver terminated early (max_iter=5000).  Consider pre-processing your data with StandardScaler or MinMaxScaler.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    }
   ],
   "source": [
    "# split df to X and Y (no train-test split. we will use all of eyeblink8 videos as training set, talkingFace as test set)\n",
    "y = ear13_df.loc[:, 'blink_annot'].values\n",
    "X = ear13_df.drop(columns=['blink_annot'])\n",
    "\n",
    "# normalize for each subject\n",
    "f = lambda x: (StandardScaler().fit_transform(x.to_frame()))[:, 0]\n",
    "norm_X = X.groupby('subject').transform(f)\n",
    "\n",
    "# normalize for all dataset\n",
    "norm_X = StandardScaler().fit_transform(norm_X)\n",
    "\n",
    "# cross-validation with 5 splits\n",
    "cv = StratifiedShuffleSplit(n_splits=10, random_state = 42)\n",
    "\n",
    "# SVM classifier\n",
    "svm = SVC(tol=1e-5)\n",
    "\n",
    "# parameters \n",
    "parameters = {\n",
    "                'kernel': ['linear', 'poly', 'rbf', 'sigmoid'],\n",
    "                'C': [0.01,0.1,1,10,100],\n",
    "                'gamma': ['scale', 'auto'],\n",
    "                #'degree': [2,3,4],\n",
    "                'max_iter': [100,1000,5000],\n",
    "                'class_weight': [None, 'balanced'],\n",
    "    \n",
    "            }\n",
    "4\n",
    "# grid search for parameters\n",
    "grid = GridSearchCV(estimator=svm, param_grid=parameters, cv=cv, n_jobs=-1)\n",
    "grid.fit(norm_X, y)\n",
    "\n",
    "# print best scores\n",
    "print(\"The best parameters are %s with a score of %0.4f\\n\"\n",
    "      % (grid.best_params_, grid.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Read test set and construct its (n x 13) matrix, then do some analysis again."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 4988 entries, 0 to 4987\n",
      "Data columns (total 15 columns):\n",
      " #   Column       Non-Null Count  Dtype  \n",
      "---  ------       --------------  -----  \n",
      " 0   subject      4988 non-null   object \n",
      " 1   blink_annot  4988 non-null   int64  \n",
      " 2   ear-6        4988 non-null   float64\n",
      " 3   ear-5        4988 non-null   float64\n",
      " 4   ear-4        4988 non-null   float64\n",
      " 5   ear-3        4988 non-null   float64\n",
      " 6   ear-2        4988 non-null   float64\n",
      " 7   ear-1        4988 non-null   float64\n",
      " 8   ear0         4988 non-null   float64\n",
      " 9   ear1         4988 non-null   float64\n",
      " 10  ear2         4988 non-null   float64\n",
      " 11  ear3         4988 non-null   float64\n",
      " 12  ear4         4988 non-null   float64\n",
      " 13  ear5         4988 non-null   float64\n",
      " 14  ear6         4988 non-null   float64\n",
      "dtypes: float64(13), int64(1), object(1)\n",
      "memory usage: 584.7+ KB\n",
      "None\n",
      "\n",
      "MISSING VALUES:\n",
      "subject        0\n",
      "blink_annot    0\n",
      "ear-6          0\n",
      "ear-5          0\n",
      "ear-4          0\n",
      "ear-3          0\n",
      "ear-2          0\n",
      "ear-1          0\n",
      "ear0           0\n",
      "ear1           0\n",
      "ear2           0\n",
      "ear3           0\n",
      "ear4           0\n",
      "ear5           0\n",
      "ear6           0\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "# read dataset\n",
    "df_test = pd.read_pickle(\"../input/eye-blink-detection-3-ml-model-part1/test/test_set.pkl\")\n",
    "\n",
    "# only get necessary columns\n",
    "ear_test = df_test[['subject','frame_no','avg_ear','blink_annot']]\n",
    "\n",
    "# group by subject\n",
    "user_list_test = list(ear_test['subject'].unique())\n",
    "list_of_dfs_test=[]\n",
    "for user in user_list_test:\n",
    "    list_of_dfs_test.append(ear_test.groupby('subject').get_group(user))\n",
    "\n",
    "\n",
    "    \n",
    "# construct a df with columns depicting 13 frame window (6 previous frames + current frame + 6 next frames)\n",
    "list_of_13dfs_test = []\n",
    "for subject_df in list_of_dfs_test:\n",
    "    for i in range(-6,7):\n",
    "        subject_df['ear'+str(i)] = subject_df.shift(periods=i*(-1))['avg_ear']\n",
    "    subject_df=subject_df[6:-6]\n",
    "    list_of_13dfs_test.append(subject_df)\n",
    "    \n",
    "# concat results\n",
    "ear13_df_test = pd.concat(list_of_13dfs_test)\n",
    "ear13_df_test.drop(columns=['frame_no', 'avg_ear'],inplace=True)\n",
    "ear13_df_test.reset_index(drop=True, inplace=True)\n",
    "\n",
    "# df info\n",
    "print(ear13_df_test.info())\n",
    "print()\n",
    "\n",
    "# Look for missing values\n",
    "print(\"MISSING VALUES:\")\n",
    "print(ear13_df_test.isnull().sum())  "
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Prepare test set for predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split df to X_test and y_test\n",
    "y_test = ear13_df_test.loc[:, 'blink_annot'].values\n",
    "X_test = ear13_df_test.drop(columns=['blink_annot'])\n",
    "\n",
    "# normalize for each subject\n",
    "f = lambda x: (StandardScaler().fit_transform(x.to_frame()))[:, 0]\n",
    "norm_X_test = X_test.groupby('subject').transform(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Implement a utility function to print results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# display test scores and return result string and indexes of false samples\n",
    "def display_test_scores_v2(test, pred):\n",
    "    str_out = \"\"\n",
    "    str_out += (\"TEST SCORES\\n\")\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    #print accuracy\n",
    "    accuracy = accuracy_score(test, pred)\n",
    "    str_out += (\"ACCURACY: {:.4f}\\n\".format(accuracy))\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    #print AUC score\n",
    "    auc = roc_auc_score(test, pred)\n",
    "    str_out += (\"AUC: {:.4f}\\n\".format(auc))\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    #print confusion matrix\n",
    "    str_out += (\"CONFUSION MATRIX:\\n\")\n",
    "    conf_mat = confusion_matrix(test, pred)\n",
    "    str_out += (\"{}\".format(conf_mat))\n",
    "    str_out += (\"\\n\")\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    #print FP, FN\n",
    "    str_out += (\"FALSE POSITIVES:\\n\")\n",
    "    fp = conf_mat[1][0]\n",
    "    pos_labels = conf_mat[1][0]+conf_mat[1][1]\n",
    "    str_out += (\"{} out of {} positive labels ({:.4f}%)\\n\".format(fp, pos_labels,fp/pos_labels))\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    str_out += (\"FALSE NEGATIVES:\\n\")\n",
    "    fn = conf_mat[0][1]\n",
    "    neg_labels = conf_mat[0][1]+conf_mat[0][0]\n",
    "    str_out += (\"{} out of {} negative labels ({:.4f}%)\\n\".format(fn, neg_labels, fn/neg_labels))\n",
    "    str_out += (\"\\n\")\n",
    "\n",
    "    #print classification report\n",
    "    str_out += (\"PRECISION, RECALL, F1 scores:\\n\")\n",
    "    str_out += (\"{}\".format(classification_report(test, pred)))\n",
    "    \n",
    "    false_indexes = np.where(test != pred)\n",
    "    return str_out, false_indexes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TEST SCORES\n",
      "\n",
      "ACCURACY: 0.9753\n",
      "\n",
      "AUC: 0.9278\n",
      "\n",
      "CONFUSION MATRIX:\n",
      "[[4471   64]\n",
      " [  59  394]]\n",
      "\n",
      "FALSE POSITIVES:\n",
      "59 out of 453 positive labels (0.1302%)\n",
      "\n",
      "FALSE NEGATIVES:\n",
      "64 out of 4535 negative labels (0.0141%)\n",
      "\n",
      "PRECISION, RECALL, F1 scores:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.99      0.99      0.99      4535\n",
      "           1       0.86      0.87      0.86       453\n",
      "\n",
      "    accuracy                           0.98      4988\n",
      "   macro avg       0.92      0.93      0.93      4988\n",
      "weighted avg       0.98      0.98      0.98      4988\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# prediction results\n",
    "y_pred = grid.predict(norm_X_test)\n",
    "\n",
    "# print accuracy metrics\n",
    "results, false = display_test_scores_v2(y_test, y_pred)\n",
    "print(results)"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "Try another SVM implementation by https://github.com/rezaghoddoosian/Early-Drowsiness-Detection and compare results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/sklearn/externals/joblib/__init__.py:15: FutureWarning: sklearn.externals.joblib is deprecated in 0.21 and will be removed in 0.23. Please import this functionality directly from joblib, which can be installed with: pip install joblib. If this warning is raised when loading pickled models, you may need to re-serialize those models with scikit-learn 0.21+.\n",
      "  warnings.warn(msg, category=FutureWarning)\n",
      "/opt/conda/lib/python3.7/site-packages/sklearn/utils/deprecation.py:144: FutureWarning: The sklearn.svm.classes module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.svm. Anything that cannot be imported from sklearn.svm is now part of the private API.\n",
      "  warnings.warn(message, FutureWarning)\n",
      "/opt/conda/lib/python3.7/site-packages/sklearn/base.py:318: UserWarning: Trying to unpickle estimator SVC from version pre-0.18 when using version 0.22.2.post1. This might lead to breaking code or invalid results. Use at your own risk.\n",
      "  UserWarning)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.externals import joblib \n",
    "\n",
    "# load the file \"Trained_SVM_C1000_gamma0.1_for 7kNegSample.sav\" \n",
    "svm2 = joblib.load(\"../input/rldd-blink-svm/Trained_SVM_C1000_gamma0.1_for 7kNegSample.sav\")  "
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "It's an old version of SVC (scikit<0.18) and its predict() method gives error so we use manual implementation of rbf prediction method based on its parameters. (see https://stackoverflow.com/questions/28593684/replication-of-scikit-svm-srv-predictx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define decision function and predict methods\n",
    "def rbf_decision_function(sup_vecs, dual_coefs, gamma, intercept,Xtest):\n",
    "    diff = sup_vecs - Xtest\n",
    "    norm2 = np.array([np.linalg.norm(diff[n, :]) for n in range(np.shape(sup_vecs)[0])])\n",
    "    dec_func_vec = (dual_coefs.dot(np.exp(-gamma*(norm2**2))) + intercept)\n",
    "    return dec_func_vec\n",
    "\n",
    "def rbf_predict(sup_vecs, dual_coefs, gamma, intercept,Xtest):\n",
    "    list_predictions=[]\n",
    "    for row in Xtest:\n",
    "        dec = rbf_decision_function(sup_vecs, dual_coefs, gamma, intercept,row)\n",
    "        list_predictions.append(1 if dec[0]>0 else 0)\n",
    "    return np.array(list_predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TEST SCORES\n",
      "\n",
      "ACCURACY: 0.9094\n",
      "\n",
      "AUC: 0.9372\n",
      "\n",
      "CONFUSION MATRIX:\n",
      "[[4096  439]\n",
      " [  13  440]]\n",
      "\n",
      "FALSE POSITIVES:\n",
      "13 out of 453 positive labels (0.0287%)\n",
      "\n",
      "FALSE NEGATIVES:\n",
      "439 out of 4535 negative labels (0.0968%)\n",
      "\n",
      "PRECISION, RECALL, F1 scores:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       1.00      0.90      0.95      4535\n",
      "           1       0.50      0.97      0.66       453\n",
      "\n",
      "    accuracy                           0.91      4988\n",
      "   macro avg       0.75      0.94      0.80      4988\n",
      "weighted avg       0.95      0.91      0.92      4988\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# get parameters from old model\n",
    "sup_vecs = svm2.support_vectors_\n",
    "dual_coefs = svm2.dual_coef_\n",
    "gamma = svm2.gamma\n",
    "intercept = svm2.intercept_/10 # ---> after some tryouts, apperantly, they did intercept scaling BY 10\n",
    "\n",
    "# prediction results\n",
    "y_pred2 = rbf_predict(sup_vecs, dual_coefs, gamma, intercept, np.array(norm_X_test))\n",
    "#y_pred2= np.apply_along_axis(rbf_decision_function,1,np.array(norm_X_test),sup_vecs, dual_coefs, gamma, intercept)\n",
    "\n",
    "# print accuracy metrics\n",
    "results2, false2 = display_test_scores_v2(y_test, (y_pred2>0).astype(int))\n",
    "print(results2)"
   ]
  },
  {
   "cell_type": "markdown",
   "execution_count": null,
   "metadata": {},
   "source": [
    "**NOTE:** Or alternatively, instead of using manual predictions, you can assign missing attributes like this:\n",
    "\n",
    "> svm2._n_support = np.array([956, 961])  \n",
    "> svm2.break_ties = False  \n",
    "\n",
    "Then you can use original predict() method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# print predictions\n",
    "pred_df=pd.Series(y_pred,name='predictions')\n",
    "pred_df.to_pickle(\"red_df_ml.pkl\")\n",
    "\n",
    "pred_df=pd.Series(y_pred2,name='predictions')\n",
    "pred_df.to_pickle(\"red_df_ml2.pkl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}