Final_Model_LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Tnr89AVpIE_0"
   },
   "source": [
    "# Preprocessing of the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 55
    },
    "colab_type": "code",
    "id": "g4K0qptsIE_2",
    "outputId": "b0d0ff3d-f682-4db8-b82a-dcbeca710947"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.colors as mcolors\n",
    "import datetime\n",
    "import sklearn\n",
    "\n",
    "#from google.colab import drive\n",
    "#drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "tYxGHSz9IE_4"
   },
   "source": [
    "Import the data for the NJ train network. Month of March 2018. Drop the observations with NaN delay."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "bpmvzos3IE_5"
   },
   "outputs": [],
   "source": [
    "#for colab : df = pd.read_csv('/content/Outward.csv')\n",
    "\n",
    "df = pd.read_csv('Outward.csv')\n",
    "df.dropna(subset=['delay_minutes'], inplace=True)\n",
    "df.reset_index(drop=True, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "0vyknETDIE_7"
   },
   "source": [
    "Encoding the days of the week (0 : Monday ..., 6 : Sunday) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LmKz3w0SIE_7"
   },
   "outputs": [],
   "source": [
    "day = []\n",
    "time = []\n",
    "\n",
    "for (k,v) in enumerate(df['actual_time']):\n",
    "    aux = datetime.datetime.strptime(v, '%Y-%m-%d %H:%M:%S')\n",
    "    time_only = aux.time()\n",
    "    day.append(aux.weekday())\n",
    "    time_in_min = time_only.hour*60 + time_only.minute\n",
    "    time.append(time_in_min)\n",
    "    \n",
    "df['day'] = day\n",
    "df['time'] = time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "ghM02ODEIE_9",
    "outputId": "fd1b1148-b1ab-4ed8-f26e-54d8042485b5"
   },
   "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>Unnamed: 0</th>\n",
       "      <th>date</th>\n",
       "      <th>train_id</th>\n",
       "      <th>stop_sequence</th>\n",
       "      <th>from</th>\n",
       "      <th>from_id</th>\n",
       "      <th>to</th>\n",
       "      <th>to_id</th>\n",
       "      <th>scheduled_time</th>\n",
       "      <th>actual_time</th>\n",
       "      <th>delay_minutes</th>\n",
       "      <th>status</th>\n",
       "      <th>line</th>\n",
       "      <th>type</th>\n",
       "      <th>day</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>3805</td>\n",
       "      <td>1.0</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>105</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>105</td>\n",
       "      <td>2018-03-02 01:22:00</td>\n",
       "      <td>2018-03-02 01:21:05</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>departed</td>\n",
       "      <td>Northeast Corrdr</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>81</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>3805</td>\n",
       "      <td>2.0</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>105</td>\n",
       "      <td>Secaucus Upper Lvl</td>\n",
       "      <td>38187</td>\n",
       "      <td>2018-03-02 01:31:00</td>\n",
       "      <td>2018-03-02 01:31:08</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>departed</td>\n",
       "      <td>Northeast Corrdr</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>3805</td>\n",
       "      <td>3.0</td>\n",
       "      <td>Secaucus Upper Lvl</td>\n",
       "      <td>38187</td>\n",
       "      <td>Newark Penn Station</td>\n",
       "      <td>107</td>\n",
       "      <td>2018-03-02 01:40:00</td>\n",
       "      <td>2018-03-02 01:40:07</td>\n",
       "      <td>0.116667</td>\n",
       "      <td>departed</td>\n",
       "      <td>Northeast Corrdr</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>3805</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Newark Penn Station</td>\n",
       "      <td>107</td>\n",
       "      <td>Newark Airport</td>\n",
       "      <td>37953</td>\n",
       "      <td>2018-03-02 01:45:00</td>\n",
       "      <td>2018-03-02 01:45:10</td>\n",
       "      <td>0.166667</td>\n",
       "      <td>departed</td>\n",
       "      <td>Northeast Corrdr</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>3805</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Newark Airport</td>\n",
       "      <td>37953</td>\n",
       "      <td>North Elizabeth</td>\n",
       "      <td>109</td>\n",
       "      <td>2018-03-02 01:49:00</td>\n",
       "      <td>2018-03-02 01:49:10</td>\n",
       "      <td>0.166667</td>\n",
       "      <td>departed</td>\n",
       "      <td>Northeast Corrdr</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>109</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0        date  train_id  stop_sequence                   from  \\\n",
       "0           0  2018-03-01      3805            1.0  New York Penn Station   \n",
       "1           1  2018-03-01      3805            2.0  New York Penn Station   \n",
       "2           2  2018-03-01      3805            3.0     Secaucus Upper Lvl   \n",
       "3           3  2018-03-01      3805            4.0    Newark Penn Station   \n",
       "4           4  2018-03-01      3805            5.0         Newark Airport   \n",
       "\n",
       "   from_id                     to  to_id       scheduled_time  \\\n",
       "0      105  New York Penn Station    105  2018-03-02 01:22:00   \n",
       "1      105     Secaucus Upper Lvl  38187  2018-03-02 01:31:00   \n",
       "2    38187    Newark Penn Station    107  2018-03-02 01:40:00   \n",
       "3      107         Newark Airport  37953  2018-03-02 01:45:00   \n",
       "4    37953        North Elizabeth    109  2018-03-02 01:49:00   \n",
       "\n",
       "           actual_time  delay_minutes    status              line        type  \\\n",
       "0  2018-03-02 01:21:05       0.000000  departed  Northeast Corrdr  NJ Transit   \n",
       "1  2018-03-02 01:31:08       0.133333  departed  Northeast Corrdr  NJ Transit   \n",
       "2  2018-03-02 01:40:07       0.116667  departed  Northeast Corrdr  NJ Transit   \n",
       "3  2018-03-02 01:45:10       0.166667  departed  Northeast Corrdr  NJ Transit   \n",
       "4  2018-03-02 01:49:10       0.166667  departed  Northeast Corrdr  NJ Transit   \n",
       "\n",
       "   day  time  \n",
       "0    4    81  \n",
       "1    4    91  \n",
       "2    4   100  \n",
       "3    4   105  \n",
       "4    4   109  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(n=5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "VjNE72GaIFAA"
   },
   "source": [
    "Encoding the line names and the departure station with label encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u4R7T4bKIFAB"
   },
   "outputs": [],
   "source": [
    "from sklearn import preprocessing\n",
    "\n",
    "le = preprocessing.LabelEncoder()\n",
    "\n",
    "df['line'] = le.fit_transform(df['line'])\n",
    "df['from_id'] = le.fit_transform(df['from_id'])\n",
    "df['train_id'] = le.fit_transform(df['train_id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 395
    },
    "colab_type": "code",
    "id": "jK-NFLV4IFAC",
    "outputId": "867ecfb5-7bdf-4302-ac65-ba48e743a96f"
   },
   "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>Unnamed: 0</th>\n",
       "      <th>date</th>\n",
       "      <th>train_id</th>\n",
       "      <th>stop_sequence</th>\n",
       "      <th>from</th>\n",
       "      <th>from_id</th>\n",
       "      <th>to</th>\n",
       "      <th>to_id</th>\n",
       "      <th>scheduled_time</th>\n",
       "      <th>actual_time</th>\n",
       "      <th>delay_minutes</th>\n",
       "      <th>status</th>\n",
       "      <th>line</th>\n",
       "      <th>type</th>\n",
       "      <th>day</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>262</td>\n",
       "      <td>1.0</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>85</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>105</td>\n",
       "      <td>2018-03-02 01:22:00</td>\n",
       "      <td>2018-03-02 01:21:05</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>departed</td>\n",
       "      <td>6</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>81</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>262</td>\n",
       "      <td>2.0</td>\n",
       "      <td>New York Penn Station</td>\n",
       "      <td>85</td>\n",
       "      <td>Secaucus Upper Lvl</td>\n",
       "      <td>38187</td>\n",
       "      <td>2018-03-02 01:31:00</td>\n",
       "      <td>2018-03-02 01:31:08</td>\n",
       "      <td>0.133333</td>\n",
       "      <td>departed</td>\n",
       "      <td>6</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>262</td>\n",
       "      <td>3.0</td>\n",
       "      <td>Secaucus Upper Lvl</td>\n",
       "      <td>143</td>\n",
       "      <td>Newark Penn Station</td>\n",
       "      <td>107</td>\n",
       "      <td>2018-03-02 01:40:00</td>\n",
       "      <td>2018-03-02 01:40:07</td>\n",
       "      <td>0.116667</td>\n",
       "      <td>departed</td>\n",
       "      <td>6</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>262</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Newark Penn Station</td>\n",
       "      <td>87</td>\n",
       "      <td>Newark Airport</td>\n",
       "      <td>37953</td>\n",
       "      <td>2018-03-02 01:45:00</td>\n",
       "      <td>2018-03-02 01:45:10</td>\n",
       "      <td>0.166667</td>\n",
       "      <td>departed</td>\n",
       "      <td>6</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>2018-03-01</td>\n",
       "      <td>262</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Newark Airport</td>\n",
       "      <td>139</td>\n",
       "      <td>North Elizabeth</td>\n",
       "      <td>109</td>\n",
       "      <td>2018-03-02 01:49:00</td>\n",
       "      <td>2018-03-02 01:49:10</td>\n",
       "      <td>0.166667</td>\n",
       "      <td>departed</td>\n",
       "      <td>6</td>\n",
       "      <td>NJ Transit</td>\n",
       "      <td>4</td>\n",
       "      <td>109</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0        date  train_id  stop_sequence                   from  \\\n",
       "0           0  2018-03-01       262            1.0  New York Penn Station   \n",
       "1           1  2018-03-01       262            2.0  New York Penn Station   \n",
       "2           2  2018-03-01       262            3.0     Secaucus Upper Lvl   \n",
       "3           3  2018-03-01       262            4.0    Newark Penn Station   \n",
       "4           4  2018-03-01       262            5.0         Newark Airport   \n",
       "\n",
       "   from_id                     to  to_id       scheduled_time  \\\n",
       "0       85  New York Penn Station    105  2018-03-02 01:22:00   \n",
       "1       85     Secaucus Upper Lvl  38187  2018-03-02 01:31:00   \n",
       "2      143    Newark Penn Station    107  2018-03-02 01:40:00   \n",
       "3       87         Newark Airport  37953  2018-03-02 01:45:00   \n",
       "4      139        North Elizabeth    109  2018-03-02 01:49:00   \n",
       "\n",
       "           actual_time  delay_minutes    status  line        type  day  time  \n",
       "0  2018-03-02 01:21:05       0.000000  departed     6  NJ Transit    4    81  \n",
       "1  2018-03-02 01:31:08       0.133333  departed     6  NJ Transit    4    91  \n",
       "2  2018-03-02 01:40:07       0.116667  departed     6  NJ Transit    4   100  \n",
       "3  2018-03-02 01:45:10       0.166667  departed     6  NJ Transit    4   105  \n",
       "4  2018-03-02 01:49:10       0.166667  departed     6  NJ Transit    4   109  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FdT9gguPIFAE"
   },
   "outputs": [],
   "source": [
    "df.drop(columns = ['date','from','to','to_id','scheduled_time','status','type','actual_time'], inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "3BZGAhJwIFAG"
   },
   "outputs": [],
   "source": [
    "df.head(n=5)\n",
    "df['delay_seconds'] = df['delay_minutes']*60\n",
    "df.drop(columns=['delay_minutes'],inplace=True)\n",
    "#len(df)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "colab_type": "code",
    "id": "ix4U53paIFAI",
    "outputId": "6afbc2d2-aade-432d-c668-69efbc4fc67d"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "19403"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.sort_values(by=['train_id','day'])\n",
    "df.head(50)\n",
    "len(df[df['line']==6])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 202
    },
    "colab_type": "code",
    "id": "Pe-4Zy-cIFAK",
    "outputId": "33a7dee4-7b04-4c9b-885d-442a7cdb4d22"
   },
   "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>Unnamed: 0</th>\n",
       "      <th>train_id</th>\n",
       "      <th>stop_sequence</th>\n",
       "      <th>from_id</th>\n",
       "      <th>line</th>\n",
       "      <th>day</th>\n",
       "      <th>time</th>\n",
       "      <th>delay_seconds</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>262</td>\n",
       "      <td>1.0</td>\n",
       "      <td>85</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>81</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>262</td>\n",
       "      <td>2.0</td>\n",
       "      <td>85</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>91</td>\n",
       "      <td>8.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>262</td>\n",
       "      <td>3.0</td>\n",
       "      <td>143</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>100</td>\n",
       "      <td>7.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>262</td>\n",
       "      <td>4.0</td>\n",
       "      <td>87</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>105</td>\n",
       "      <td>10.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>262</td>\n",
       "      <td>5.0</td>\n",
       "      <td>139</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>109</td>\n",
       "      <td>10.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0  train_id  stop_sequence  from_id  line  day  time  \\\n",
       "0           0       262            1.0       85     6    4    81   \n",
       "1           1       262            2.0       85     6    4    91   \n",
       "2           2       262            3.0      143     6    4   100   \n",
       "3           3       262            4.0       87     6    4   105   \n",
       "4           4       262            5.0      139     6    4   109   \n",
       "\n",
       "   delay_seconds  \n",
       "0            0.0  \n",
       "1            8.0  \n",
       "2            7.0  \n",
       "3           10.0  \n",
       "4           10.0  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Jbi3ltjR9t0Z"
   },
   "source": [
    "Trial on one line, for one stop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 171
    },
    "colab_type": "code",
    "id": "mKGopVdR6Y9H",
    "outputId": "f716f5e0-5197-45cc-a4ef-4be73f60f8a6"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#The create_dataset function is greatly inspired by : \n",
    "#https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/\n",
    "\n",
    "# LSTM with window regression framing\n",
    "\n",
    "import numpy\n",
    "import matplotlib.pyplot as plt\n",
    "from pandas import read_csv\n",
    "import math\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.layers import LSTM\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "#for colab : \n",
    "#import tensorflow.compat.v1 as tf\n",
    "#tf.disable_v2_behavior()\n",
    "\n",
    "# convert an array of values into a dataset matrix\n",
    "def create_dataset(dataset, look_back=1):\n",
    "    dataX, dataY = [], []\n",
    "    for i in range(len(dataset)-look_back-1):\n",
    "        a = dataset[i:(i+look_back), :] #it can take as many features as we want \n",
    "        dataX.append(a)\n",
    "        dataY.append(dataset[i + look_back, 0])\n",
    "    return numpy.array(dataX), numpy.array(dataY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "8QZenmbf7xHo"
   },
   "outputs": [],
   "source": [
    "# fix random seed for reproducibility\n",
    "numpy.random.seed(7)\n",
    "\n",
    "# load the dataset, I chose to work on line 6 only as an example\n",
    "df_line_6 = df[df['line']==6]\n",
    "dataset = df_line_6[['delay_seconds','time']] #adding delays and actual time as a feature\n",
    "\n",
    "# normalize the dataset\n",
    "scaler = MinMaxScaler(feature_range=(0, 1))\n",
    "dataset = scaler.fit_transform(dataset)\n",
    "\n",
    "dataset_test = dataset[df_line_6['stop_sequence'] == 10]\n",
    "dataset_train = dataset[df_line_6['stop_sequence'] != 10]\n",
    "\n",
    "#dataset_test = dataset_test.values\n",
    "dataset_test = dataset_test.astype('float32')\n",
    "\n",
    "#dataset_train = dataset_train.values\n",
    "dataset_train = dataset_train.astype('float32')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 69
    },
    "colab_type": "code",
    "id": "jvBrtySKUrOZ",
    "outputId": "2b59277b-48b7-499a-f020-3c883941c588"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(19403, 2)\n",
      "(18270, 2)\n",
      "(1133, 2)\n"
     ]
    }
   ],
   "source": [
    "#Sanity check\n",
    "\n",
    "print(dataset.shape)\n",
    "print(dataset_train.shape)\n",
    "print(dataset_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "WjWEfN8FNbWA"
   },
   "outputs": [],
   "source": [
    "# split into train and test sets\n",
    "train = dataset_train\n",
    "test = dataset_test\n",
    "\n",
    "# reshape into X=t and Y=t+look_back-1\n",
    "look_back = 4 #for example\n",
    "trainX, trainY = create_dataset(train, look_back)\n",
    "testX, testY = create_dataset(test, look_back)\n",
    "\n",
    "# reshape input dimensions to be [delay, time, features] with the features being the previous delays and time\n",
    "trainX = numpy.reshape(trainX, (trainX.shape[0], trainX.shape[1], trainX.shape[2]))\n",
    "testX = numpy.reshape(testX, (testX.shape[0], testX.shape[1], testX.shape[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 52
    },
    "colab_type": "code",
    "id": "oqCRwIr-YkRP",
    "outputId": "978a7d15-b682-4c7d-d240-55d0d64b8728"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1128, 4, 2)\n",
      "(1128,)\n"
     ]
    }
   ],
   "source": [
    "#Sanity check\n",
    "\n",
    "print(testX.shape)\n",
    "print(testY.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "G1igqXKWNfmT",
    "outputId": "50be1893-2293-42f5-a801-270d07c05fb8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      " - 35s - loss: 0.0032\n",
      "Epoch 2/10\n",
      " - 36s - loss: 0.0027\n",
      "Epoch 3/10\n",
      " - 38s - loss: 0.0027\n",
      "Epoch 4/10\n",
      " - 43s - loss: 0.0027\n",
      "Epoch 5/10\n",
      " - 42s - loss: 0.0027\n",
      "Epoch 6/10\n",
      " - 39s - loss: 0.0027\n",
      "Epoch 7/10\n",
      " - 37s - loss: 0.0026\n",
      "Epoch 8/10\n",
      " - 37s - loss: 0.0027\n",
      "Epoch 9/10\n",
      " - 37s - loss: 0.0026\n",
      "Epoch 10/10\n",
      " - 38s - loss: 0.0026\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_1 (LSTM)                (None, 10)                520       \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 11        \n",
      "=================================================================\n",
      "Total params: 531\n",
      "Trainable params: 531\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# create and fit the LSTM network\n",
    "model = Sequential()\n",
    "model.add(LSTM(10, input_shape=(look_back, 2)))\n",
    "model.add(Dense(1))\n",
    "model.compile(loss='mean_squared_error', optimizer='adam')\n",
    "model.fit(trainX, trainY, epochs=10, batch_size=1, verbose=2)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "-hPn4T5PNkRN"
   },
   "outputs": [],
   "source": [
    "# make predictions\n",
    "trainPredict = model.predict(trainX)\n",
    "testPredict = model.predict(testX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "otYZx_9iuD8m"
   },
   "outputs": [],
   "source": [
    "#for now, manual inverse scaling, so we should get the min and max for each set\n",
    "\n",
    "df_stop_test = df_line_6[df_line_6['stop_sequence']==10]\n",
    "df_groundtruth_test = df_stop_test['delay_seconds']\n",
    "\n",
    "df_stop_train = df_line_6[df_line_6['stop_sequence']!=10]\n",
    "df_groundtruth_train = df_stop_train['delay_seconds']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 52
    },
    "colab_type": "code",
    "id": "N2Cx3cLRxR-y",
    "outputId": "cc0844cb-fb6d-45ec-8a44-b1149b56a1fc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5100.0\n",
      "5305.000000002\n"
     ]
    }
   ],
   "source": [
    "print(np.max(df_groundtruth_test))\n",
    "print(np.max(df_groundtruth_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "lBMDpGN8NopB"
   },
   "outputs": [],
   "source": [
    "#Inverse scaling\n",
    "\n",
    "trainPredict_norm = trainPredict*5305\n",
    "testPredict_norm = testPredict*5305\n",
    "trainY_norm = trainY*5305\n",
    "testY_norm = testY*5305"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 52
    },
    "colab_type": "code",
    "id": "kM76QaEQNrT9",
    "outputId": "e0b746ef-2d68-41c1-b9e4-58f957015a56"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Score in minutes: 4.50 RMSE\n",
      "Test Score in minutes: 10.13 RMSE\n"
     ]
    }
   ],
   "source": [
    "# calculate root mean squared error\n",
    "trainScore = math.sqrt(mean_squared_error(trainY_norm, trainPredict_norm[:,0]))\n",
    "print('Train Score in minutes: %.2f RMSE' % (trainScore/60))\n",
    "testScore = math.sqrt(mean_squared_error(testY_norm, testPredict_norm[:,0]))\n",
    "print('Test Score in minutes: %.2f RMSE' % (testScore/60))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 202
    },
    "colab_type": "code",
    "id": "nfEnNcH9sB2V",
    "outputId": "a2976293-7cb9-4fb1-adba-6c03bcdc2b6b"
   },
   "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>Time [hour]</th>\n",
       "      <th>TestY [min]</th>\n",
       "      <th>Predict [min]</th>\n",
       "      <th>Groundtruth [min]</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>511</td>\n",
       "      <td>16.633333</td>\n",
       "      <td>1.092206</td>\n",
       "      <td>47.163326</td>\n",
       "      <td>1.050000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>582</td>\n",
       "      <td>17.366667</td>\n",
       "      <td>1.248235</td>\n",
       "      <td>3.472070</td>\n",
       "      <td>1.200000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>945</td>\n",
       "      <td>10.483333</td>\n",
       "      <td>36.701588</td>\n",
       "      <td>1.963092</td>\n",
       "      <td>35.283333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1018</td>\n",
       "      <td>9.016667</td>\n",
       "      <td>1.300245</td>\n",
       "      <td>29.474165</td>\n",
       "      <td>1.250000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1331</td>\n",
       "      <td>22.816667</td>\n",
       "      <td>2.236421</td>\n",
       "      <td>3.013185</td>\n",
       "      <td>2.150000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      Time [hour]  TestY [min]  Predict [min]  Groundtruth [min]\n",
       "511     16.633333     1.092206      47.163326           1.050000\n",
       "582     17.366667     1.248235       3.472070           1.200000\n",
       "945     10.483333    36.701588       1.963092          35.283333\n",
       "1018     9.016667     1.300245      29.474165           1.250000\n",
       "1331    22.816667     2.236421       3.013185           2.150000"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Transform back to delays in minutes, the time still has to change too\n",
    "\n",
    "groundtruth_test = df_groundtruth_test.to_numpy()\n",
    "groundtruth_train = df_groundtruth_train.to_numpy()\n",
    "\n",
    "df_time = df['time'][(df['line'] == 6) & (df['stop_sequence']==10)]/60\n",
    "np.max(df_time)\n",
    "\n",
    "df_test_plot = pd.DataFrame()\n",
    "df_test_plot['Time [hour]'] = df_time[4:-1] \n",
    "df_test_plot['TestY [min]'] = testY_norm*5305/5100/60\n",
    "df_test_plot['Predict [min]'] = testPredict_norm/60\n",
    "df_test_plot['Groundtruth [min]'] = groundtruth_test[4:-1]/60 #attention to this, caused by the look back\n",
    "\n",
    "df_test_plot.head() #sanity check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 281
    },
    "colab_type": "code",
    "id": "8zR4evYwNzy1",
    "outputId": "aac68475-3842-4ee9-e8dd-f5c43dab7bc8"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax1 = plt.scatter(df_test_plot['Time [hour]'],df_test_plot['Groundtruth [min]'],label='Groundtruth')\n",
    "ax2 = plt.scatter(df_test_plot['Time [hour]'],df_test_plot['Predict [min]'],label='Prediction')\n",
    "#plt.title('Prediction of delays at stop \"Metropark\", on line \"Northeast Corridor\" for outward trips ')\n",
    "plt.ylabel('Delay [min]')\n",
    "plt.xlabel('Time of the day')\n",
    "plt.legend()\n",
    "plt.savefig('./Images and Gifs/LSTM_plot.pdf')\n",
    "\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Swcx29qnhCEG"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "Final_Model_LSTM.ipynb",
   "provenance": []
  },
  "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}