Jyoti_Autoencoder_anomaly.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Jyoti_Autoencoder_anomaly.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/nrjcs/iitpbse/blob/master/Jyoti_Autoencoder_anomaly.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mmHZT_pJN4O2",
        "colab_type": "code",
        "colab": {
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/files.js": {
              "data": "Ly8gQ29weXJpZ2h0IDIwMTcgR29vZ2xlIExMQwovLwovLyBMaWNlbnNlZCB1bmRlciB0aGUgQXBhY2hlIExpY2Vuc2UsIFZlcnNpb24gMi4wICh0aGUgIkxpY2Vuc2UiKTsKLy8geW91IG1heSBub3QgdXNlIHRoaXMgZmlsZSBleGNlcHQgaW4gY29tcGxpYW5jZSB3aXRoIHRoZSBMaWNlbnNlLgovLyBZb3UgbWF5IG9idGFpbiBhIGNvcHkgb2YgdGhlIExpY2Vuc2UgYXQKLy8KLy8gICAgICBodHRwOi8vd3d3LmFwYWNoZS5vcmcvbGljZW5zZXMvTElDRU5TRS0yLjAKLy8KLy8gVW5sZXNzIHJlcXVpcmVkIGJ5IGFwcGxpY2FibGUgbGF3IG9yIGFncmVlZCB0byBpbiB3cml0aW5nLCBzb2Z0d2FyZQovLyBkaXN0cmlidXRlZCB1bmRlciB0aGUgTGljZW5zZSBpcyBkaXN0cmlidXRlZCBvbiBhbiAiQVMgSVMiIEJBU0lTLAovLyBXSVRIT1VUIFdBUlJBTlRJRVMgT1IgQ09ORElUSU9OUyBPRiBBTlkgS0lORCwgZWl0aGVyIGV4cHJlc3Mgb3IgaW1wbGllZC4KLy8gU2VlIHRoZSBMaWNlbnNlIGZvciB0aGUgc3BlY2lmaWMgbGFuZ3VhZ2UgZ292ZXJuaW5nIHBlcm1pc3Npb25zIGFuZAovLyBsaW1pdGF0aW9ucyB1bmRlciB0aGUgTGljZW5zZS4KCi8qKgogKiBAZmlsZW92ZXJ2aWV3IEhlbHBlcnMgZm9yIGdvb2dsZS5jb2xhYiBQeXRob24gbW9kdWxlLgogKi8KKGZ1bmN0aW9uKHNjb3BlKSB7CmZ1bmN0aW9uIHNwYW4odGV4dCwgc3R5bGVBdHRyaWJ1dGVzID0ge30pIHsKICBjb25zdCBlbGVtZW50ID0gZG9jdW1lbnQuY3JlYXRlRWxlbWVudCgnc3BhbicpOwogIGVsZW1lbnQudGV4dENvbnRlbnQgPSB0ZXh0OwogIGZvciAoY29uc3Qga2V5IG9mIE9iamVjdC5rZXlzKHN0eWxlQXR0cmlidXRlcykpIHsKICAgIGVsZW1lbnQuc3R5bGVba2V5XSA9IHN0eWxlQXR0cmlidXRlc1trZXldOwogIH0KICByZXR1cm4gZWxlbWVudDsKfQoKLy8gTWF4IG51bWJlciBvZiBieXRlcyB3aGljaCB3aWxsIGJlIHVwbG9hZGVkIGF0IGEgdGltZS4KY29uc3QgTUFYX1BBWUxPQURfU0laRSA9IDEwMCAqIDEwMjQ7CgpmdW5jdGlvbiBfdXBsb2FkRmlsZXMoaW5wdXRJZCwgb3V0cHV0SWQpIHsKICBjb25zdCBzdGVwcyA9IHVwbG9hZEZpbGVzU3RlcChpbnB1dElkLCBvdXRwdXRJZCk7CiAgY29uc3Qgb3V0cHV0RWxlbWVudCA9IGRvY3VtZW50LmdldEVsZW1lbnRCeUlkKG91dHB1dElkKTsKICAvLyBDYWNoZSBzdGVwcyBvbiB0aGUgb3V0cHV0RWxlbWVudCB0byBtYWtlIGl0IGF2YWlsYWJsZSBmb3IgdGhlIG5leHQgY2FsbAogIC8vIHRvIHVwbG9hZEZpbGVzQ29udGludWUgZnJvbSBQeXRob24uCiAgb3V0cHV0RWxlbWVudC5zdGVwcyA9IHN0ZXBzOwoKICByZXR1cm4gX3VwbG9hZEZpbGVzQ29udGludWUob3V0cHV0SWQpOwp9CgovLyBUaGlzIGlzIHJvdWdobHkgYW4gYXN5bmMgZ2VuZXJhdG9yIChub3Qgc3VwcG9ydGVkIGluIHRoZSBicm93c2VyIHlldCksCi8vIHdoZXJlIHRoZXJlIGFyZSBtdWx0aXBsZSBhc3luY2hyb25vdXMgc3RlcHMgYW5kIHRoZSBQeXRob24gc2lkZSBpcyBnb2luZwovLyB0byBwb2xsIGZvciBjb21wbGV0aW9uIG9mIGVhY2ggc3RlcC4KLy8gVGhpcyB1c2VzIGEgUHJvbWlzZSB0byBibG9jayB0aGUgcHl0aG9uIHNpZGUgb24gY29tcGxldGlvbiBvZiBlYWNoIHN0ZXAsCi8vIHRoZW4gcGFzc2VzIHRoZSByZXN1bHQgb2YgdGhlIHByZXZpb3VzIHN0ZXAgYXMgdGhlIGlucHV0IHRvIHRoZSBuZXh0IHN0ZXAuCmZ1bmN0aW9uIF91cGxvYWRGaWxlc0NvbnRpbnVlKG91dHB1dElkKSB7CiAgY29uc3Qgb3V0cHV0RWxlbWVudCA9IGRvY3VtZW50LmdldEVsZW1lbnRCeUlkKG91dHB1dElkKTsKICBjb25zdCBzdGVwcyA9IG91dHB1dEVsZW1lbnQuc3RlcHM7CgogIGNvbnN0IG5leHQgPSBzdGVwcy5uZXh0KG91dHB1dEVsZW1lbnQubGFzdFByb21pc2VWYWx1ZSk7CiAgcmV0dXJuIFByb21pc2UucmVzb2x2ZShuZXh0LnZhbHVlLnByb21pc2UpLnRoZW4oKHZhbHVlKSA9PiB7CiAgICAvLyBDYWNoZSB0aGUgbGFzdCBwcm9taXNlIHZhbHVlIHRvIG1ha2UgaXQgYXZhaWxhYmxlIHRvIHRoZSBuZXh0CiAgICAvLyBzdGVwIG9mIHRoZSBnZW5lcmF0b3IuCiAgICBvdXRwdXRFbGVtZW50Lmxhc3RQcm9taXNlVmFsdWUgPSB2YWx1ZTsKICAgIHJldHVybiBuZXh0LnZhbHVlLnJlc3BvbnNlOwogIH0pOwp9CgovKioKICogR2VuZXJhdG9yIGZ1bmN0aW9uIHdoaWNoIGlzIGNhbGxlZCBiZXR3ZWVuIGVhY2ggYXN5bmMgc3RlcCBvZiB0aGUgdXBsb2FkCiAqIHByb2Nlc3MuCiAqIEBwYXJhbSB7c3RyaW5nfSBpbnB1dElkIEVsZW1lbnQgSUQgb2YgdGhlIGlucHV0IGZpbGUgcGlja2VyIGVsZW1lbnQuCiAqIEBwYXJhbSB7c3RyaW5nfSBvdXRwdXRJZCBFbGVtZW50IElEIG9mIHRoZSBvdXRwdXQgZGlzcGxheS4KICogQHJldHVybiB7IUl0ZXJhYmxlPCFPYmplY3Q+fSBJdGVyYWJsZSBvZiBuZXh0IHN0ZXBzLgogKi8KZnVuY3Rpb24qIHVwbG9hZEZpbGVzU3RlcChpbnB1dElkLCBvdXRwdXRJZCkgewogIGNvbnN0IGlucHV0RWxlbWVudCA9IGRvY3VtZW50LmdldEVsZW1lbnRCeUlkKGlucHV0SWQpOwogIGlucHV0RWxlbWVudC5kaXNhYmxlZCA9IGZhbHNlOwoKICBjb25zdCBvdXRwdXRFbGVtZW50ID0gZG9jdW1lbnQuZ2V0RWxlbWVudEJ5SWQob3V0cHV0SWQpOwogIG91dHB1dEVsZW1lbnQuaW5uZXJIVE1MID0gJyc7CgogIGNvbnN0IHBpY2tlZFByb21pc2UgPSBuZXcgUHJvbWlzZSgocmVzb2x2ZSkgPT4gewogICAgaW5wdXRFbGVtZW50LmFkZEV2ZW50TGlzdGVuZXIoJ2NoYW5nZScsIChlKSA9PiB7CiAgICAgIHJlc29sdmUoZS50YXJnZXQuZmlsZXMpOwogICAgfSk7CiAgfSk7CgogIGNvbnN0IGNhbmNlbCA9IGRvY3VtZW50LmNyZWF0ZUVsZW1lbnQoJ2J1dHRvbicpOwogIGlucHV0RWxlbWVudC5wYXJlbnRFbGVtZW50LmFwcGVuZENoaWxkKGNhbmNlbCk7CiAgY2FuY2VsLnRleHRDb250ZW50ID0gJ0NhbmNlbCB1cGxvYWQnOwogIGNvbnN0IGNhbmNlbFByb21pc2UgPSBuZXcgUHJvbWlzZSgocmVzb2x2ZSkgPT4gewogICAgY2FuY2VsLm9uY2xpY2sgPSAoKSA9PiB7CiAgICAgIHJlc29sdmUobnVsbCk7CiAgICB9OwogIH0pOwoKICAvLyBXYWl0IGZvciB0aGUgdXNlciB0byBwaWNrIHRoZSBmaWxlcy4KICBjb25zdCBmaWxlcyA9IHlpZWxkIHsKICAgIHByb21pc2U6IFByb21pc2UucmFjZShbcGlja2VkUHJvbWlzZSwgY2FuY2VsUHJvbWlzZV0pLAogICAgcmVzcG9uc2U6IHsKICAgICAgYWN0aW9uOiAnc3RhcnRpbmcnLAogICAgfQogIH07CgogIGNhbmNlbC5yZW1vdmUoKTsKCiAgLy8gRGlzYWJsZSB0aGUgaW5wdXQgZWxlbWVudCBzaW5jZSBmdXJ0aGVyIHBpY2tzIGFyZSBub3QgYWxsb3dlZC4KICBpbnB1dEVsZW1lbnQuZGlzYWJsZWQgPSB0cnVlOwoKICBpZiAoIWZpbGVzKSB7CiAgICByZXR1cm4gewogICAgICByZXNwb25zZTogewogICAgICAgIGFjdGlvbjogJ2NvbXBsZXRlJywKICAgICAgfQogICAgfTsKICB9CgogIGZvciAoY29uc3QgZmlsZSBvZiBmaWxlcykgewogICAgY29uc3QgbGkgPSBkb2N1bWVudC5jcmVhdGVFbGVtZW50KCdsaScpOwogICAgbGkuYXBwZW5kKHNwYW4oZmlsZS5uYW1lLCB7Zm9udFdlaWdodDogJ2JvbGQnfSkpOwogICAgbGkuYXBwZW5kKHNwYW4oCiAgICAgICAgYCgke2ZpbGUudHlwZSB8fCAnbi9hJ30pIC0gJHtmaWxlLnNpemV9IGJ5dGVzLCBgICsKICAgICAgICBgbGFzdCBtb2RpZmllZDogJHsKICAgICAgICAgICAgZmlsZS5sYXN0TW9kaWZpZWREYXRlID8gZmlsZS5sYXN0TW9kaWZpZWREYXRlLnRvTG9jYWxlRGF0ZVN0cmluZygpIDoKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgJ24vYSd9IC0gYCkpOwogICAgY29uc3QgcGVyY2VudCA9IHNwYW4oJzAlIGRvbmUnKTsKICAgIGxpLmFwcGVuZENoaWxkKHBlcmNlbnQpOwoKICAgIG91dHB1dEVsZW1lbnQuYXBwZW5kQ2hpbGQobGkpOwoKICAgIGNvbnN0IGZpbGVEYXRhUHJvbWlzZSA9IG5ldyBQcm9taXNlKChyZXNvbHZlKSA9PiB7CiAgICAgIGNvbnN0IHJlYWRlciA9IG5ldyBGaWxlUmVhZGVyKCk7CiAgICAgIHJlYWRlci5vbmxvYWQgPSAoZSkgPT4gewogICAgICAgIHJlc29sdmUoZS50YXJnZXQucmVzdWx0KTsKICAgICAgfTsKICAgICAgcmVhZGVyLnJlYWRBc0FycmF5QnVmZmVyKGZpbGUpOwogICAgfSk7CiAgICAvLyBXYWl0IGZvciB0aGUgZGF0YSB0byBiZSByZWFkeS4KICAgIGxldCBmaWxlRGF0YSA9IHlpZWxkIHsKICAgICAgcHJvbWlzZTogZmlsZURhdGFQcm9taXNlLAogICAgICByZXNwb25zZTogewogICAgICAgIGFjdGlvbjogJ2NvbnRpbnVlJywKICAgICAgfQogICAgfTsKCiAgICAvLyBVc2UgYSBjaHVua2VkIHNlbmRpbmcgdG8gYXZvaWQgbWVzc2FnZSBzaXplIGxpbWl0cy4gU2VlIGIvNjIxMTU2NjAuCiAgICBsZXQgcG9zaXRpb24gPSAwOwogICAgd2hpbGUgKHBvc2l0aW9uIDwgZmlsZURhdGEuYnl0ZUxlbmd0aCkgewogICAgICBjb25zdCBsZW5ndGggPSBNYXRoLm1pbihmaWxlRGF0YS5ieXRlTGVuZ3RoIC0gcG9zaXRpb24sIE1BWF9QQVlMT0FEX1NJWkUpOwogICAgICBjb25zdCBjaHVuayA9IG5ldyBVaW50OEFycmF5KGZpbGVEYXRhLCBwb3NpdGlvbiwgbGVuZ3RoKTsKICAgICAgcG9zaXRpb24gKz0gbGVuZ3RoOwoKICAgICAgY29uc3QgYmFzZTY0ID0gYnRvYShTdHJpbmcuZnJvbUNoYXJDb2RlLmFwcGx5KG51bGwsIGNodW5rKSk7CiAgICAgIHlpZWxkIHsKICAgICAgICByZXNwb25zZTogewogICAgICAgICAgYWN0aW9uOiAnYXBwZW5kJywKICAgICAgICAgIGZpbGU6IGZpbGUubmFtZSwKICAgICAgICAgIGRhdGE6IGJhc2U2NCwKICAgICAgICB9LAogICAgICB9OwogICAgICBwZXJjZW50LnRleHRDb250ZW50ID0KICAgICAgICAgIGAke01hdGgucm91bmQoKHBvc2l0aW9uIC8gZmlsZURhdGEuYnl0ZUxlbmd0aCkgKiAxMDApfSUgZG9uZWA7CiAgICB9CiAgfQoKICAvLyBBbGwgZG9uZS4KICB5aWVsZCB7CiAgICByZXNwb25zZTogewogICAgICBhY3Rpb246ICdjb21wbGV0ZScsCiAgICB9CiAgfTsKfQoKc2NvcGUuZ29vZ2xlID0gc2NvcGUuZ29vZ2xlIHx8IHt9OwpzY29wZS5nb29nbGUuY29sYWIgPSBzY29wZS5nb29nbGUuY29sYWIgfHwge307CnNjb3BlLmdvb2dsZS5jb2xhYi5fZmlsZXMgPSB7CiAgX3VwbG9hZEZpbGVzLAogIF91cGxvYWRGaWxlc0NvbnRpbnVlLAp9Owp9KShzZWxmKTsK",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "status": 200,
              "status_text": ""
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 73
        },
        "outputId": "d994820b-2123-4428-90c2-2f7a8dbde199"
      },
      "source": [
        "from google.colab import files\n",
        "uploaded = files.upload()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-4a669519-c7ee-4d55-bc62-06ca813d15cf\" name=\"files[]\" multiple disabled\n",
              "        style=\"border:none\" />\n",
              "     <output id=\"result-4a669519-c7ee-4d55-bc62-06ca813d15cf\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script src=\"/nbextensions/google.colab/files.js\"></script> "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Saving data1.csv to data1.csv\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1yka4HcdWLDi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        },
        "outputId": "2779a9b4-8834-4164-b611-486b149817e3"
      },
      "source": [
        "import pandas as pd\n",
        "#https://pandas.pydata.org/docs/reference/index.html\n",
        "\n",
        "import numpy as np\n",
        "#https://numpy.org/doc/stable/user/tutorials_index.html\n",
        "\n",
        "from matplotlib import pyplot as plt\n",
        "#https://matplotlib.org/tutorials/index.html\n",
        "\n",
        "from keras.callbacks import ModelCheckpoint, TensorBoard\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.model_selection import train_test_split\n",
        "#https://scikit-learn.org/stable/user_guide.html\n",
        "\n",
        "from keras.layers import Input, Dense\n",
        "from keras.models import Model\n",
        "from keras import regularizers\n",
        "#https://keras.io/api/\n",
        "\n",
        "import seaborn as sns\n",
        "#https://seaborn.pydata.org/api.html"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/statsmodels/tools/_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n",
            "  import pandas.util.testing as tm\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BdOfzucxWbw7",
        "colab_type": "text"
      },
      "source": [
        "**Anomaly detection using autoencoder**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hk_bXbEpUClZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df = pd.read_csv(\"data1.csv\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RzifaSBsW06N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data2= df.drop(['timestamp','time_epoch'], axis=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OFWllShJvnJN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "41ae0323-18cc-4e43-9435-5beb527b9f3c"
      },
      "source": [
        "print(data2[7030:7040])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "          value  hours  daylight  DayOfTheWeek  WeekDay  categories  anomaly\n",
            "7030  15.077814     19         1             6        0           1        0\n",
            "7031  15.186548     20         1             6        0           1        0\n",
            "7032  14.724430     21         1             6        0           1        1\n",
            "7033  15.059782     22         1             6        0           1        1\n",
            "7034  14.533523     23         0             6        0           0        1\n",
            "7035  14.679799      0         0             0        1           2        1\n",
            "7036  14.367725      1         0             0        1           2        1\n",
            "7037  14.799608      2         0             0        1           2        0\n",
            "7038  15.041495      3         0             0        1           2        0\n",
            "7039  15.828292      4         0             0        1           2        0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1z6Xt2JFZTpG",
        "colab_type": "text"
      },
      "source": [
        "# Split the data in to training, test, and validation set."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LusqgOqnYu1S",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df2_train, df2_test = train_test_split(data2, test_size=0.2,random_state=123 )\n",
        "df2_train, df2_valid = train_test_split(df2_train, test_size=0.2,random_state=123)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kx79eZH5b_Sa",
        "colab_type": "text"
      },
      "source": [
        "# Remove the anomalous instances from training and validation set.\n",
        "\n",
        "**Basic steps to Anomaly Detection using an Autoencoder:**\n",
        "\n",
        " - Train an Autoencoder on normal data (no anomalies)\n",
        "\n",
        "\n",
        " - Take a new data point and try to reconstruct it using the Autoencoder\n",
        "\n",
        " - If the error (reconstruction error) for the new data point is above some threshold, we label the example as an anomaly."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4Cg7G9QRcC2p",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_data21 = df2_train.loc[df2_train['anomaly'] == 0, ['value', 'hours', 'daylight', 'DayOfTheWeek', 'WeekDay','categories','anomaly']]\n",
        "val_data21 = df2_valid.loc[df2_valid['anomaly'] == 0, ['value', 'hours', 'daylight', 'DayOfTheWeek','WeekDay','categories', 'anomaly']]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CqLIaJ_yckKB",
        "colab_type": "text"
      },
      "source": [
        "**# Drop the label column**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hDOa987WcO-9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "train_data2= train_data21.drop(['anomaly'], axis=1)\n",
        "val_data2=val_data21.drop(['anomaly'], axis=1)\n",
        "test_data2=df2_test.drop(['anomaly'], axis=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QuxoRSBOdSuy",
        "colab_type": "text"
      },
      "source": [
        "**# Size of the train, validation and test data set**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aeWQKLGqcuT_",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        },
        "outputId": "71ae48bf-51f7-4f19-e45e-119b1d293af2"
      },
      "source": [
        "print(train_data2.shape)\n",
        "print(val_data2.shape)\n",
        "print(test_data2.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(4608, 6)\n",
            "(1151, 6)\n",
            "(1454, 6)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FlgVj9D0dzGD",
        "colab_type": "text"
      },
      "source": [
        "**# parameters**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tAC-bi1MdQU1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 64\n",
        "input_dim = 6 \n",
        "encoding_dim = 64\n",
        "hidden_dim = int(encoding_dim / 2)\n",
        "learning_rate = 1e-3"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u80mcuiJeM9F",
        "colab_type": "text"
      },
      "source": [
        "# Create the autoencoder model \n",
        "\n",
        "**Autoencoder :**\n",
        "\n",
        " - Autoencoder takes the input and encodes it into a hidden representation. (encoder) \\\n",
        "\n",
        " - It decodes the input again from this hidden representation. (decoder) \\\n",
        "\n",
        "-   The model is trained to minimize a loss function (MSE) to ensure that Xi is close to X'i \n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fl1DRI5ueLGo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "f59341fd-a8fe-4d11-f45f-2f5f485e4992"
      },
      "source": [
        "input_layer = Input(shape=(input_dim, ))\n",
        "encoder = Dense(encoding_dim, activation=\"relu\", activity_regularizer=regularizers.l1(learning_rate))(input_layer)\n",
        "encoder = Dense(hidden_dim, activation=\"relu\")(encoder)\n",
        "decoder = Dense(hidden_dim, activation=\"relu\")(encoder)\n",
        "decoder = Dense(encoding_dim, activation=\"relu\")(decoder)\n",
        "decoder = Dense(input_dim, activation=\"linear\")(decoder)\n",
        "autoencoder = Model(inputs=input_layer, outputs=decoder)\n",
        "autoencoder.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"functional_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         [(None, 6)]               0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 64)                448       \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 32)                2080      \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 32)                1056      \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 64)                2112      \n",
            "_________________________________________________________________\n",
            "dense_4 (Dense)              (None, 6)                 390       \n",
            "=================================================================\n",
            "Total params: 6,086\n",
            "Trainable params: 6,086\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1iaad-5Xe-vH",
        "colab_type": "text"
      },
      "source": [
        "**# Compile and train the model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZyMbHx8LfHJO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 785
        },
        "outputId": "15f7bd3f-fd2d-4a43-95aa-c5ae703abe30"
      },
      "source": [
        "autoencoder.compile(metrics=['accuracy'],\n",
        "                    loss='mean_squared_error',\n",
        "                    optimizer='adam')\n",
        "\n",
        "cp = ModelCheckpoint(filepath=\"autoencoder_classifier.h5\",\n",
        "                               save_best_only=True,\n",
        "                               verbose=0)\n",
        "\n",
        "tb = TensorBoard(log_dir='./logs',\n",
        "                histogram_freq=0,\n",
        "                write_graph=True,\n",
        "                write_images=True)\n",
        "\n",
        "history = autoencoder.fit(train_data2, train_data2,\n",
        "                    epochs=20,\n",
        "                    batch_size=batch_size,\n",
        "                    shuffle=True,\n",
        "                    validation_data=(val_data2, val_data2),\n",
        "                    verbose=1,\n",
        "                    callbacks=[cp, tb]).history"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            " 1/72 [..............................] - ETA: 0s - loss: 102.4097 - accuracy: 0.9531WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/summary_ops_v2.py:1277: stop (from tensorflow.python.eager.profiler) is deprecated and will be removed after 2020-07-01.\n",
            "Instructions for updating:\n",
            "use `tf.profiler.experimental.stop` instead.\n",
            "WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0086s vs `on_train_batch_end` time: 0.0384s). Check your callbacks.\n",
            "72/72 [==============================] - 0s 5ms/step - loss: 24.3428 - accuracy: 0.9336 - val_loss: 2.2895 - val_accuracy: 0.9409\n",
            "Epoch 2/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.9867 - accuracy: 0.9835 - val_loss: 0.4473 - val_accuracy: 0.9922\n",
            "Epoch 3/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.2829 - accuracy: 0.9939 - val_loss: 0.2277 - val_accuracy: 0.9922\n",
            "Epoch 4/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.2179 - accuracy: 0.9963 - val_loss: 0.2088 - val_accuracy: 0.9974\n",
            "Epoch 5/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1997 - accuracy: 0.9961 - val_loss: 0.1905 - val_accuracy: 0.9957\n",
            "Epoch 6/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1842 - accuracy: 0.9957 - val_loss: 0.1761 - val_accuracy: 0.9974\n",
            "Epoch 7/20\n",
            "72/72 [==============================] - 0s 3ms/step - loss: 0.1705 - accuracy: 0.9946 - val_loss: 0.1589 - val_accuracy: 0.9948\n",
            "Epoch 8/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1539 - accuracy: 0.9959 - val_loss: 0.1474 - val_accuracy: 0.9957\n",
            "Epoch 9/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1446 - accuracy: 0.9963 - val_loss: 0.1404 - val_accuracy: 0.9965\n",
            "Epoch 10/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1389 - accuracy: 0.9970 - val_loss: 0.1368 - val_accuracy: 0.9983\n",
            "Epoch 11/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1348 - accuracy: 0.9965 - val_loss: 0.1323 - val_accuracy: 0.9974\n",
            "Epoch 12/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1317 - accuracy: 0.9967 - val_loss: 0.1297 - val_accuracy: 0.9974\n",
            "Epoch 13/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1285 - accuracy: 0.9970 - val_loss: 0.1263 - val_accuracy: 0.9965\n",
            "Epoch 14/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1255 - accuracy: 0.9963 - val_loss: 0.1239 - val_accuracy: 0.9974\n",
            "Epoch 15/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1226 - accuracy: 0.9970 - val_loss: 0.1210 - val_accuracy: 0.9974\n",
            "Epoch 16/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1199 - accuracy: 0.9963 - val_loss: 0.1181 - val_accuracy: 0.9974\n",
            "Epoch 17/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1171 - accuracy: 0.9970 - val_loss: 0.1151 - val_accuracy: 0.9974\n",
            "Epoch 18/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1141 - accuracy: 0.9967 - val_loss: 0.1125 - val_accuracy: 0.9974\n",
            "Epoch 19/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1111 - accuracy: 0.9970 - val_loss: 0.1099 - val_accuracy: 0.9965\n",
            "Epoch 20/20\n",
            "72/72 [==============================] - 0s 2ms/step - loss: 0.1085 - accuracy: 0.9974 - val_loss: 0.1064 - val_accuracy: 0.9974\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EGpV4K2KfzS8",
        "colab_type": "text"
      },
      "source": [
        "**# training loss vs validation loss**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fRr-Yp54fVF0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "8a64303a-a589-409c-fcd4-f9881959dbb3"
      },
      "source": [
        "plt.plot(history['loss'], linewidth=2, label='Train')\n",
        "plt.plot(history['val_loss'], linewidth=2, label='Valid')\n",
        "plt.legend(loc='upper right')\n",
        "plt.title('Model loss')\n",
        "plt.ylabel('Loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z5glBvmKgEW1",
        "colab_type": "text"
      },
      "source": [
        "**# Test the autoencoder model on test dataset**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pDwq2Z89fpjC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_x_predictions = autoencoder.predict(test_data2)\n",
        "t=test_data2.values\n",
        "mse = np.mean(np.power(t - test_x_predictions, 2), axis=1)  \n",
        "error_df_test = pd.DataFrame({'Reconstruction_error': mse,\n",
        "                        'True_class': df2_test['anomaly']})\n",
        "error_df_test = error_df_test.reset_index()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nknDA22EgxE0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "82ad51d6-64cf-4e84-e855-717a5347c6ff"
      },
      "source": [
        "threshold_fixed = 0.006\n",
        "groups = error_df_test.groupby('True_class')\n",
        "\n",
        "fig, ax = plt.subplots()\n",
        "\n",
        "for name, group in groups:\n",
        "    ax.plot(group.index, group.Reconstruction_error, marker='o', ms=4.5, linestyle='',\n",
        "            label= \"Break\" if name == 1 else \"Normal\")\n",
        "ax.hlines(threshold_fixed, ax.get_xlim()[0], ax.get_xlim()[1], colors=\"r\", zorder=100, label='Threshold')\n",
        "ax.legend()\n",
        "plt.title(\"Reconstruction error for different classes\")\n",
        "plt.ylabel(\"Reconstruction error\")\n",
        "plt.xlabel(\"Data point index\")\n",
        "plt.show();"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f0mO_feRhOQK",
        "colab_type": "text"
      },
      "source": [
        "**# Confusion matrix**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bq5IhEOBg04u",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 350
        },
        "outputId": "27be1ea5-9bde-4aaf-9d5c-6e4293b86411"
      },
      "source": [
        "pred_y = [1 if e > threshold_fixed else 0 for e in error_df_test.Reconstruction_error.values]\n",
        "conf_matrix = confusion_matrix(error_df_test.True_class, pred_y)\n",
        "plt.figure(figsize=(5, 5))\n",
        "sns.heatmap(conf_matrix, annot=True, fmt=\"d\");\n",
        "plt.title(\"Confusion matrix\")\n",
        "plt.ylabel('True class')\n",
        "plt.xlabel('Predicted class')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}