Encoder Decoder for Time series forecasting.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled0.ipynb",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/debaonline4u/Machine-Learning-with-Python/blob/main/Encoder%20Decoder%20for%20Time%20series%20forecasting.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "w7C2_c_iauR0"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from sklearn import preprocessing\n",
        "import matplotlib.pyplot as plt\n",
        "tf.random.set_seed(123)\n",
        "np.random.seed(123)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "url = f\"https://raw.githubusercontent.com/Apress/hands-on-time-series-analylsis-python/master/Data/Metro_Interstate_Traffic_Volume.csv\""
      ],
      "metadata": {
        "id": "f1zW-TvRi19P"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "data = pd.read_csv(url)"
      ],
      "metadata": {
        "id": "5J5U8WnfcS7i"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "data.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 400
        },
        "id": "XxxLLA6xdVGd",
        "outputId": "a2749575-6143-4c0a-a2ea-6919868ebb06"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>holiday</th>\n",
              "      <th>temp</th>\n",
              "      <th>rain_1h</th>\n",
              "      <th>snow_1h</th>\n",
              "      <th>clouds_all</th>\n",
              "      <th>weather_main</th>\n",
              "      <th>weather_description</th>\n",
              "      <th>date_time</th>\n",
              "      <th>traffic_volume</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>None</td>\n",
              "      <td>288.28</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>40</td>\n",
              "      <td>Clouds</td>\n",
              "      <td>scattered clouds</td>\n",
              "      <td>2012-10-02 09:00:00</td>\n",
              "      <td>5545</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>None</td>\n",
              "      <td>289.36</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>75</td>\n",
              "      <td>Clouds</td>\n",
              "      <td>broken clouds</td>\n",
              "      <td>2012-10-02 10:00:00</td>\n",
              "      <td>4516</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>None</td>\n",
              "      <td>289.58</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>90</td>\n",
              "      <td>Clouds</td>\n",
              "      <td>overcast clouds</td>\n",
              "      <td>2012-10-02 11:00:00</td>\n",
              "      <td>4767</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>None</td>\n",
              "      <td>290.13</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>90</td>\n",
              "      <td>Clouds</td>\n",
              "      <td>overcast clouds</td>\n",
              "      <td>2012-10-02 12:00:00</td>\n",
              "      <td>5026</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>None</td>\n",
              "      <td>291.14</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>75</td>\n",
              "      <td>Clouds</td>\n",
              "      <td>broken clouds</td>\n",
              "      <td>2012-10-02 13:00:00</td>\n",
              "      <td>4918</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  holiday    temp  ...            date_time  traffic_volume\n",
              "0    None  288.28  ...  2012-10-02 09:00:00            5545\n",
              "1    None  289.36  ...  2012-10-02 10:00:00            4516\n",
              "2    None  289.58  ...  2012-10-02 11:00:00            4767\n",
              "3    None  290.13  ...  2012-10-02 12:00:00            5026\n",
              "4    None  291.14  ...  2012-10-02 13:00:00            4918\n",
              "\n",
              "[5 rows x 9 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "data.describe()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "mLY-Ca7nmkV-",
        "outputId": "c99b768a-182a-4065-d42b-4030dd76911b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>temp</th>\n",
              "      <th>rain_1h</th>\n",
              "      <th>snow_1h</th>\n",
              "      <th>clouds_all</th>\n",
              "      <th>traffic_volume</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>48204.000000</td>\n",
              "      <td>48204.000000</td>\n",
              "      <td>48204.000000</td>\n",
              "      <td>48204.000000</td>\n",
              "      <td>48204.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>281.205870</td>\n",
              "      <td>0.334264</td>\n",
              "      <td>0.000222</td>\n",
              "      <td>49.362231</td>\n",
              "      <td>3259.818355</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>13.338232</td>\n",
              "      <td>44.789133</td>\n",
              "      <td>0.008168</td>\n",
              "      <td>39.015750</td>\n",
              "      <td>1986.860670</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>272.160000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>1.000000</td>\n",
              "      <td>1193.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>282.450000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>64.000000</td>\n",
              "      <td>3380.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>291.806000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>90.000000</td>\n",
              "      <td>4933.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>310.070000</td>\n",
              "      <td>9831.300000</td>\n",
              "      <td>0.510000</td>\n",
              "      <td>100.000000</td>\n",
              "      <td>7280.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               temp       rain_1h       snow_1h    clouds_all  traffic_volume\n",
              "count  48204.000000  48204.000000  48204.000000  48204.000000    48204.000000\n",
              "mean     281.205870      0.334264      0.000222     49.362231     3259.818355\n",
              "std       13.338232     44.789133      0.008168     39.015750     1986.860670\n",
              "min        0.000000      0.000000      0.000000      0.000000        0.000000\n",
              "25%      272.160000      0.000000      0.000000      1.000000     1193.000000\n",
              "50%      282.450000      0.000000      0.000000     64.000000     3380.000000\n",
              "75%      291.806000      0.000000      0.000000     90.000000     4933.000000\n",
              "max      310.070000   9831.300000      0.510000    100.000000     7280.000000"
            ]
          },
          "metadata": {},
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "data.drop_duplicates(subset=['date_time'], keep=False,inplace=True)"
      ],
      "metadata": {
        "id": "u9iP489Cm8US"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "validate = data['traffic_volume'].tail(10)\n",
        "data = data.drop(data['traffic_volume'].tail(10).index)"
      ],
      "metadata": {
        "id": "mrqn44F5nVyu"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "uni_data = data['traffic_volume']\n",
        "uni_data.index = data['date_time']\n",
        "uni_data.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pun1yWTepSRt",
        "outputId": "3c34e146-dc06-407a-fe8e-5a95313ddf60"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "date_time\n",
              "2012-10-02 09:00:00    5545\n",
              "2012-10-02 10:00:00    4516\n",
              "2012-10-02 11:00:00    4767\n",
              "2012-10-02 12:00:00    5026\n",
              "2012-10-02 13:00:00    4918\n",
              "Name: traffic_volume, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "uni_data = uni_data.values\n",
        "scaler_x = preprocessing.MinMaxScaler()\n",
        "x_rescaled = scaler_x.fit_transform(uni_data.reshape(-1, 1))"
      ],
      "metadata": {
        "id": "tknWNmLbp8oR"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def custom_ts_univariate_data_prep(dataset, start, end, window, horizon):\n",
        "  X = []\n",
        "  y = []\n",
        "\n",
        "  start = start + window\n",
        "  if end is None:\n",
        "    end = len(dataset) - horizon\n",
        "\n",
        "  for i in range(start, end):\n",
        "    indicesx = range(i-window, i)\n",
        "    X.append(np.reshape(dataset[indicesx], (window, 1)))\n",
        "    indicesy = range(i,i+horizon)\n",
        "    y.append(dataset[indicesy])\n",
        "  return np.array(X), np.array(y)"
      ],
      "metadata": {
        "id": "QVHeZXptqA46"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "univar_hist_window = 48\n",
        "horizon = 10\n",
        "TRAIN_SPLIT = 30000\n",
        "x_train_uni, y_train_uni = custom_ts_univariate_data_prep(x_rescaled, 0, TRAIN_SPLIT,univar_hist_window, horizon)\n",
        "x_val_uni, y_val_uni = custom_ts_univariate_data_prep(x_rescaled, TRAIN_SPLIT, None,univar_hist_window,horizon)"
      ],
      "metadata": {
        "id": "87qHiXfKqYov"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print ('Single window of past history')\n",
        "print (x_train_uni[0])\n",
        "print ('\\n Target horizon')\n",
        "print (y_train_uni[0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ayrlaT7BqhUb",
        "outputId": "2039f95d-79a4-4b8d-8dde-01949fc00515"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Single window of past history\n",
            "[[0.76167582]\n",
            " [0.62032967]\n",
            " [0.65480769]\n",
            " [0.69038462]\n",
            " [0.67554945]\n",
            " [0.71167582]\n",
            " [0.76703297]\n",
            " [0.82623626]\n",
            " [0.79546703]\n",
            " [0.65521978]\n",
            " [0.48612637]\n",
            " [0.38241758]\n",
            " [0.32431319]\n",
            " [0.21002747]\n",
            " [0.13228022]\n",
            " [0.06950549]\n",
            " [0.04409341]\n",
            " [0.0375    ]\n",
            " [0.05041209]\n",
            " [0.11181319]\n",
            " [0.37335165]\n",
            " [0.77925824]\n",
            " [0.89436813]\n",
            " [0.75151099]\n",
            " [0.70013736]\n",
            " [0.67129121]\n",
            " [0.7331044 ]\n",
            " [0.78186813]\n",
            " [0.84299451]\n",
            " [0.63502747]\n",
            " [0.49326923]\n",
            " [0.39807692]\n",
            " [0.36222527]\n",
            " [0.24409341]\n",
            " [0.13942308]\n",
            " [0.08214286]\n",
            " [0.05068681]\n",
            " [0.04285714]\n",
            " [0.05041209]\n",
            " [0.1146978 ]\n",
            " [0.37445055]\n",
            " [0.78145604]\n",
            " [0.96016484]\n",
            " [0.82211538]\n",
            " [0.72925824]\n",
            " [0.63228022]\n",
            " [0.67087912]\n",
            " [0.7010989 ]]\n",
            "\n",
            " Target horizon\n",
            "[[0.71126374]\n",
            " [0.75563187]\n",
            " [0.78475275]\n",
            " [0.86428571]\n",
            " [0.83200549]\n",
            " [0.67403846]\n",
            " [0.48118132]\n",
            " [0.41717033]\n",
            " [0.38763736]\n",
            " [0.27362637]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "BATCH_SIZE = 256\n",
        "BUFFER_SIZE = 150\n",
        "train_univariate = tf.data.Dataset.from_tensor_slices((x_train_uni, y_train_uni))\n",
        "train_univariate = train_univariate.cache().shuffle(BUFFER_SIZE).batch(BATCH_SIZE).repeat()\n",
        "\n",
        "val_univariate = tf.data.Dataset.from_tensor_slices((x_val_uni, y_val_uni))\n",
        "val_univariate = val_univariate.batch(BATCH_SIZE).repeat()"
      ],
      "metadata": {
        "id": "wak7xGrqqt2k"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = tf.keras.models.Sequential([\n",
        "    tf.keras.layers.LSTM(100, input_shape=x_train_uni.shape[-2:], return_sequences=True),\n",
        "    tf.keras.layers.LSTM(units=50,return_sequences=True),\n",
        "    tf.keras.layers.LSTM(units=15),\n",
        "    tf.keras.layers.RepeatVector(y_train_uni.shape[1]), \n",
        "    tf.keras.layers.LSTM(units=100,return_sequences=True),\n",
        "    tf.keras.layers.LSTM(units=50,return_sequences=True),\n",
        "    tf.keras.layers.TimeDistributed(tf.keras.layers.Dense(units=1))\n",
        "])\n",
        "ED_lstm_model.compile(optimizer='adam', loss='mse')"
      ],
      "metadata": {
        "id": "Eu40jBcaq1qf"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "enco_deco.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cvIF9x_Kq9Dj",
        "outputId": "63c1ba13-02e9-40e6-9961-a2df992d5a99"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " lstm (LSTM)                 (None, 48, 100)           40800     \n",
            "                                                                 \n",
            " lstm_1 (LSTM)               (None, 48, 50)            30200     \n",
            "                                                                 \n",
            " lstm_2 (LSTM)               (None, 15)                3960      \n",
            "                                                                 \n",
            " repeat_vector (RepeatVector  (None, 10, 15)           0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " lstm_3 (LSTM)               (None, 10, 100)           46400     \n",
            "                                                                 \n",
            " lstm_4 (LSTM)               (None, 10, 50)            30200     \n",
            "                                                                 \n",
            " time_distributed (TimeDistr  (None, 10, 1)            51        \n",
            " ibuted)                                                         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 151,611\n",
            "Trainable params: 151,611\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "EVALUATION_INTERVAL = 100\n",
        "EPOCHS = 150\n",
        "history = ED_lstm_model.fit(train_univariate, epochs=150,steps_per_epoch=100,validation_data=val_univariate, validation_steps=50,verbose =1)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fXUgJQjfrdqW",
        "outputId": "5cd93d8b-5070-4c2b-d5d8-d751ed90c3ec"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/150\n",
            "100/100 [==============================] - 55s 450ms/step - loss: 0.0958 - val_loss: 0.0709\n",
            "Epoch 2/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0605 - val_loss: 0.0573\n",
            "Epoch 3/150\n",
            "100/100 [==============================] - 43s 428ms/step - loss: 0.0537 - val_loss: 0.0753\n",
            "Epoch 4/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0491 - val_loss: 0.0461\n",
            "Epoch 5/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0499 - val_loss: 0.0424\n",
            "Epoch 6/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0455 - val_loss: 0.0418\n",
            "Epoch 7/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0441 - val_loss: 0.0359\n",
            "Epoch 8/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0436 - val_loss: 0.0357\n",
            "Epoch 9/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0418 - val_loss: 0.0461\n",
            "Epoch 10/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0383 - val_loss: 0.0629\n",
            "Epoch 11/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0361 - val_loss: 0.0329\n",
            "Epoch 12/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0381 - val_loss: 0.0337\n",
            "Epoch 13/150\n",
            "100/100 [==============================] - 43s 433ms/step - loss: 0.0351 - val_loss: 0.0333\n",
            "Epoch 14/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0351 - val_loss: 0.0284\n",
            "Epoch 15/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0350 - val_loss: 0.0298\n",
            "Epoch 16/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0341 - val_loss: 0.0353\n",
            "Epoch 17/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0311 - val_loss: 0.0426\n",
            "Epoch 18/150\n",
            "100/100 [==============================] - 43s 429ms/step - loss: 0.0304 - val_loss: 0.0301\n",
            "Epoch 19/150\n",
            "100/100 [==============================] - 43s 429ms/step - loss: 0.0328 - val_loss: 0.0272\n",
            "Epoch 20/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0299 - val_loss: 0.0308\n",
            "Epoch 21/150\n",
            "100/100 [==============================] - 43s 428ms/step - loss: 0.0308 - val_loss: 0.0262\n",
            "Epoch 22/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0307 - val_loss: 0.0259\n",
            "Epoch 23/150\n",
            "100/100 [==============================] - 43s 429ms/step - loss: 0.0304 - val_loss: 0.0289\n",
            "Epoch 24/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0277 - val_loss: 0.0308\n",
            "Epoch 25/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0279 - val_loss: 0.0258\n",
            "Epoch 26/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0290 - val_loss: 0.0275\n",
            "Epoch 27/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0275 - val_loss: 0.0263\n",
            "Epoch 28/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0287 - val_loss: 0.0250\n",
            "Epoch 29/150\n",
            "100/100 [==============================] - 43s 430ms/step - loss: 0.0283 - val_loss: 0.0259\n",
            "Epoch 30/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0285 - val_loss: 0.0258\n",
            "Epoch 31/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0254 - val_loss: 0.0256\n",
            "Epoch 32/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0268 - val_loss: 0.0303\n",
            "Epoch 33/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0266 - val_loss: 0.0250\n",
            "Epoch 34/150\n",
            "100/100 [==============================] - ETA: 0s - loss: 0.0262Epoch 35/150\n",
            "100/100 [==============================] - 43s 427ms/step - loss: 0.0274 - val_loss: 0.0250\n",
            "Epoch 36/150\n",
            "100/100 [==============================] - 43s 428ms/step - loss: 0.0269 - val_loss: 0.0268\n",
            "Epoch 37/150\n",
            "100/100 [==============================] - 43s 431ms/step - loss: 0.0271 - val_loss: 0.0256\n",
            "Epoch 38/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0238 - val_loss: 0.0256\n",
            "Epoch 39/150\n",
            "100/100 [==============================] - 43s 433ms/step - loss: 0.0260 - val_loss: 0.0305\n",
            "Epoch 40/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0253 - val_loss: 0.0238\n",
            "Epoch 41/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0249 - val_loss: 0.0258\n",
            "Epoch 42/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0259 - val_loss: 0.0249\n",
            "Epoch 43/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0254 - val_loss: 0.0260\n",
            "Epoch 44/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0259 - val_loss: 0.0277\n",
            "Epoch 45/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0225 - val_loss: 0.0266\n",
            "Epoch 46/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0254 - val_loss: 0.0276\n",
            "Epoch 47/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0245 - val_loss: 0.0242\n",
            "Epoch 48/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0239 - val_loss: 0.0252\n",
            "Epoch 49/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0253 - val_loss: 0.0233\n",
            "Epoch 50/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0247 - val_loss: 0.0256\n",
            "Epoch 51/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0252 - val_loss: 0.0328\n",
            "Epoch 52/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0217 - val_loss: 0.0257\n",
            "Epoch 53/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0247 - val_loss: 0.0236\n",
            "Epoch 54/150\n",
            "100/100 [==============================] - 45s 447ms/step - loss: 0.0230 - val_loss: 0.0245\n",
            "Epoch 55/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0230 - val_loss: 0.0231\n",
            "Epoch 56/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0240 - val_loss: 0.0222\n",
            "Epoch 57/150\n",
            "100/100 [==============================] - 43s 436ms/step - loss: 0.0243 - val_loss: 0.0259\n",
            "Epoch 58/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0249 - val_loss: 0.0344\n",
            "Epoch 59/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0214 - val_loss: 0.0239\n",
            "Epoch 60/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0241 - val_loss: 0.0242\n",
            "Epoch 61/150\n",
            "100/100 [==============================] - 44s 444ms/step - loss: 0.0224 - val_loss: 0.0251\n",
            "Epoch 62/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0224 - val_loss: 0.0222\n",
            "Epoch 63/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0232 - val_loss: 0.0227\n",
            "Epoch 64/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0228 - val_loss: 0.0262\n",
            "Epoch 65/150\n",
            "100/100 [==============================] - 44s 443ms/step - loss: 0.0233 - val_loss: 0.0324\n",
            "Epoch 66/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0207 - val_loss: 0.0239\n",
            "Epoch 67/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0234 - val_loss: 0.0254\n",
            "Epoch 68/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0214 - val_loss: 0.0245\n",
            "Epoch 69/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0218 - val_loss: 0.0231\n",
            "Epoch 70/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0227 - val_loss: 0.0237\n",
            "Epoch 71/150\n",
            "100/100 [==============================] - 43s 433ms/step - loss: 0.0225 - val_loss: 0.0260\n",
            "Epoch 72/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0218 - val_loss: 0.0296\n",
            "Epoch 73/150\n",
            "100/100 [==============================] - 43s 433ms/step - loss: 0.0206 - val_loss: 0.0244\n",
            "Epoch 74/150\n",
            "100/100 [==============================] - 43s 433ms/step - loss: 0.0225 - val_loss: 0.0262\n",
            "Epoch 75/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0216 - val_loss: 0.0254\n",
            "Epoch 76/150\n",
            "100/100 [==============================] - 43s 432ms/step - loss: 0.0215 - val_loss: 0.0238\n",
            "Epoch 77/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0217 - val_loss: 0.0244\n",
            "Epoch 78/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0217 - val_loss: 0.0268\n",
            "Epoch 79/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0205 - val_loss: 0.0257\n",
            "Epoch 80/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0197 - val_loss: 0.0254\n",
            "Epoch 81/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0212 - val_loss: 0.0259\n",
            "Epoch 82/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0195 - val_loss: 0.0251\n",
            "Epoch 83/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0205 - val_loss: 0.0244\n",
            "Epoch 84/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0206 - val_loss: 0.0254\n",
            "Epoch 85/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0208 - val_loss: 0.0281\n",
            "Epoch 86/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0197 - val_loss: 0.0252\n",
            "Epoch 87/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0195 - val_loss: 0.0248\n",
            "Epoch 88/150\n",
            "100/100 [==============================] - 44s 445ms/step - loss: 0.0200 - val_loss: 0.0262\n",
            "Epoch 89/150\n",
            "100/100 [==============================] - 44s 443ms/step - loss: 0.0192 - val_loss: 0.0246\n",
            "Epoch 90/150\n",
            "100/100 [==============================] - 45s 446ms/step - loss: 0.0196 - val_loss: 0.0240\n",
            "Epoch 91/150\n",
            "100/100 [==============================] - 44s 444ms/step - loss: 0.0193 - val_loss: 0.0260\n",
            "Epoch 92/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0197 - val_loss: 0.0284\n",
            "Epoch 93/150\n",
            "100/100 [==============================] - 44s 443ms/step - loss: 0.0181 - val_loss: 0.0237\n",
            "Epoch 94/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0190 - val_loss: 0.0239\n",
            "Epoch 95/150\n",
            "100/100 [==============================] - 44s 445ms/step - loss: 0.0187 - val_loss: 0.0255\n",
            "Epoch 96/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0184 - val_loss: 0.0243\n",
            "Epoch 97/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0191 - val_loss: 0.0250\n",
            "Epoch 98/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0186 - val_loss: 0.0271\n",
            "Epoch 99/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0194 - val_loss: 0.0302\n",
            "Epoch 100/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0174 - val_loss: 0.0236\n",
            "Epoch 101/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0188 - val_loss: 0.0267\n",
            "Epoch 102/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0182 - val_loss: 0.0246\n",
            "Epoch 103/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0178 - val_loss: 0.0242\n",
            "Epoch 104/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0183 - val_loss: 0.0254\n",
            "Epoch 105/150\n",
            "100/100 [==============================] - 44s 444ms/step - loss: 0.0183 - val_loss: 0.0284\n",
            "Epoch 106/150\n",
            "100/100 [==============================] - 44s 444ms/step - loss: 0.0190 - val_loss: 0.0310\n",
            "Epoch 107/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0167 - val_loss: 0.0247\n",
            "Epoch 108/150\n",
            "100/100 [==============================] - 45s 446ms/step - loss: 0.0181 - val_loss: 0.0286\n",
            "Epoch 109/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0176 - val_loss: 0.0248\n",
            "Epoch 110/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0174 - val_loss: 0.0246\n",
            "Epoch 111/150\n",
            "100/100 [==============================] - 44s 443ms/step - loss: 0.0175 - val_loss: 0.0253\n",
            "Epoch 112/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0173 - val_loss: 0.0294\n",
            "Epoch 113/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0179 - val_loss: 0.0274\n",
            "Epoch 114/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0161 - val_loss: 0.0265\n",
            "Epoch 115/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0176 - val_loss: 0.0285\n",
            "Epoch 116/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0163 - val_loss: 0.0253\n",
            "Epoch 117/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0165 - val_loss: 0.0242\n",
            "Epoch 118/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0166 - val_loss: 0.0252\n",
            "Epoch 119/150\n",
            "100/100 [==============================] - 43s 434ms/step - loss: 0.0165 - val_loss: 0.0310\n",
            "Epoch 120/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0172 - val_loss: 0.0267\n",
            "Epoch 121/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0150 - val_loss: 0.0280\n",
            "Epoch 122/150\n",
            "100/100 [==============================] - 44s 444ms/step - loss: 0.0163 - val_loss: 0.0294\n",
            "Epoch 123/150\n",
            "100/100 [==============================] - 45s 446ms/step - loss: 0.0154 - val_loss: 0.0265\n",
            "Epoch 124/150\n",
            "100/100 [==============================] - 44s 436ms/step - loss: 0.0159 - val_loss: 0.0251\n",
            "Epoch 125/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0160 - val_loss: 0.0264\n",
            "Epoch 126/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0160 - val_loss: 0.0301\n",
            "Epoch 127/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0166 - val_loss: 0.0268\n",
            "Epoch 128/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0148 - val_loss: 0.0278\n",
            "Epoch 129/150\n",
            "100/100 [==============================] - 45s 447ms/step - loss: 0.0161 - val_loss: 0.0295\n",
            "Epoch 130/150\n",
            "100/100 [==============================] - 45s 449ms/step - loss: 0.0151 - val_loss: 0.0265\n",
            "Epoch 131/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0153 - val_loss: 0.0261\n",
            "Epoch 132/150\n",
            "100/100 [==============================] - 44s 441ms/step - loss: 0.0150 - val_loss: 0.0283\n",
            "Epoch 133/150\n",
            "100/100 [==============================] - 44s 446ms/step - loss: 0.0151 - val_loss: 0.0281\n",
            "Epoch 134/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0152 - val_loss: 0.0290\n",
            "Epoch 135/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0144 - val_loss: 0.0277\n",
            "Epoch 136/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0156 - val_loss: 0.0287\n",
            "Epoch 137/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0140 - val_loss: 0.0277\n",
            "Epoch 138/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0141 - val_loss: 0.0282\n",
            "Epoch 139/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0146 - val_loss: 0.0275\n",
            "Epoch 140/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0146 - val_loss: 0.0289\n",
            "Epoch 141/150\n",
            "100/100 [==============================] - 44s 442ms/step - loss: 0.0141 - val_loss: 0.0272\n",
            "Epoch 142/150\n",
            "100/100 [==============================] - 44s 437ms/step - loss: 0.0136 - val_loss: 0.0292\n",
            "Epoch 143/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0146 - val_loss: 0.0274\n",
            "Epoch 144/150\n",
            "100/100 [==============================] - 44s 439ms/step - loss: 0.0139 - val_loss: 0.0273\n",
            "Epoch 145/150\n",
            "100/100 [==============================] - 44s 443ms/step - loss: 0.0144 - val_loss: 0.0285\n",
            "Epoch 146/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0138 - val_loss: 0.0286\n",
            "Epoch 147/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0139 - val_loss: 0.0303\n",
            "Epoch 148/150\n",
            "100/100 [==============================] - 44s 438ms/step - loss: 0.0132 - val_loss: 0.0289\n",
            "Epoch 149/150\n",
            "100/100 [==============================] - 44s 440ms/step - loss: 0.0132 - val_loss: 0.0283\n",
            "Epoch 150/150\n",
            "100/100 [==============================] - 43s 435ms/step - loss: 0.0133 - val_loss: 0.0271\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# create model\n",
        "from tensorflow import keras\n",
        "from keras.models import Sequential\n",
        "from keras import layers\n",
        "from keras.layers import LSTM, TimeDistributed\n",
        "enco_deco = Sequential()\n",
        "# Encoder\n",
        "enco_deco.add(LSTM(100, input_shape=x_train_uni.shape[-2:], return_sequences=True))\n",
        "enco_deco.add(LSTM(units=50,return_sequences=True))\n",
        "enco_deco.add(LSTM(units=15))\n",
        "#feature vector\n",
        "enco_deco.add(layers.RepeatVector(y_train_uni.shape[1]))\n",
        "#decoder\n",
        "enco_deco.add(LSTM(units=100,return_sequences=True))\n",
        "enco_deco.add(LSTM(units=50,return_sequences=True))\n",
        "enco_deco.add(TimeDistributed(tf.keras.layers.Dense(units=1)))"
      ],
      "metadata": {
        "id": "VEmY67Bt3byF"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "enco_deco.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Sz9g7yy9Fn7e",
        "outputId": "5892f5f3-1494-4587-da5d-e68db604b2fd"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " lstm_13 (LSTM)              (None, 48, 100)           40800     \n",
            "                                                                 \n",
            " lstm_14 (LSTM)              (None, 48, 50)            30200     \n",
            "                                                                 \n",
            " lstm_15 (LSTM)              (None, 15)                3960      \n",
            "                                                                 \n",
            " repeat_vector_2 (RepeatVect  (None, 10, 15)           0         \n",
            " or)                                                             \n",
            "                                                                 \n",
            " lstm_16 (LSTM)              (None, 10, 100)           46400     \n",
            "                                                                 \n",
            " lstm_17 (LSTM)              (None, 10, 50)            30200     \n",
            "                                                                 \n",
            " time_distributed_1 (TimeDis  (None, 10, 1)            51        \n",
            " tributed)                                                       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 151,611\n",
            "Trainable params: 151,611\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "uni = data['traffic_volume']\n",
        "validatehori = uni.tail(48)"
      ],
      "metadata": {
        "id": "Q8jboYHiF2Nl"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "validatehist = validatehori.values\n",
        "scaler_val = preprocessing.MinMaxScaler()\n",
        "val_rescaled = scaler_x.fit_transform(validatehist.reshape(-1, 1))"
      ],
      "metadata": {
        "id": "EAEQcrUPF-24"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "val_rescaled = val_rescaled.reshape((1, val_rescaled.shape[0], 1))"
      ],
      "metadata": {
        "id": "iO8VJs18GA45"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Predicted_results = ED_lstm_model.predict(val_rescaled)"
      ],
      "metadata": {
        "id": "_RBuo6X0GDSj"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Predicted_results"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uy1PVpaEGKRD",
        "outputId": "d570a060-0bc8-48ee-db34-dab9ed581794"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[[0.6388559 ],\n",
              "        [0.63769186],\n",
              "        [0.63238335],\n",
              "        [0.6240387 ],\n",
              "        [0.60370827],\n",
              "        [0.5555512 ],\n",
              "        [0.4813918 ],\n",
              "        [0.40486968],\n",
              "        [0.33662346],\n",
              "        [0.27134418]]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 56
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "Predicted_inver_res = scaler_x.inverse_transform(Predicted_results[0])"
      ],
      "metadata": {
        "id": "DjIvp4H5GTYd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "Predicted_inver_res"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EhvKk2EpGWsM",
        "outputId": "ce417f1b-c226-4ff1-d56f-969326d7c549"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[4074.2234],\n",
              "       [4067.4104],\n",
              "       [4036.3396],\n",
              "       [3987.4985],\n",
              "       [3868.5044],\n",
              "       [3586.6409],\n",
              "       [3152.5862],\n",
              "       [2704.7021],\n",
              "       [2305.257 ],\n",
              "       [1923.1775]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 58
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn import metrics\n",
        "def timeseries_evaluation_metrics_func(y_true, y_pred):\n",
        "    \n",
        "    def mean_absolute_percentage_error(y_true, y_pred): \n",
        "        y_true, y_pred = np.array(y_true), np.array(y_pred)\n",
        "        return np.mean(np.abs((y_true - y_pred) / y_true)) * 100\n",
        "    print('Evaluation metric results:-')\n",
        "    print(f'MSE is : {metrics.mean_squared_error(validate,Predicted_inver_res)}')\n",
        "    print(f'MAE is : {metrics.mean_absolute_error(validate,Predicted_inver_res)}')\n",
        "    print(f'RMSE is : {np.sqrt(metrics.mean_squared_error(validate,Predicted_inver_res))}')\n",
        "    print(f'MAPE is : {mean_absolute_percentage_error(validate,Predicted_inver_res)}')\n",
        "    print(f'R2 is : {metrics.r2_score(validate,Predicted_inver_res)}',end='\\n\\n')"
      ],
      "metadata": {
        "id": "p4k0XVx9GZ0L"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "timeseries_evaluation_metrics_func(validate,Predicted_inver_res)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Nu70OsCiGdRJ",
        "outputId": "32745070-6c68-468a-f93c-67f63d4cf68d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Evaluation metric results:-\n",
            "MSE is : 255229.7545795381\n",
            "MAE is : 408.9387451171875\n",
            "RMSE is : 505.202686631354\n",
            "MAPE is : 59.963241499333805\n",
            "R2 is : 0.8343010016255124\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.plot( list(validate))\n",
        "plt.plot( list(Predicted_inver_res))\n",
        "plt.title(\"Actual vs Predicted\")\n",
        "plt.ylabel(\"Traffic volume\")\n",
        "plt.legend(('Actual','predicted'))\n",
        "plt.rcParams[\"figure.figsize\"] = [16,9]\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 488
        },
        "id": "HxnIvJ-2GnUd",
        "outputId": "7a8c63b0-b0ae-40de-cce9-8c8bde1f2e76"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1152x648 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}