Anomaly_detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Anomaly_detection.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "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/KristynaPijackova/Tutorials_NNs_and_signals/blob/main/Anomaly_detection.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Anomaly detection with ECG dataset"
      ],
      "metadata": {
        "id": "lrp9P-gZWijt"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "So in our previous task we used the ECG5000 dataset for a classification task. \n",
        "\n",
        "However, as we noticed when analyzing the dataset, the classes are really imbalanced and thus not suitable for classification task. We tried to improve this by creating a synthetic data with SMOTE method (Synthetic Minority Oversampling TEchnique). While this helped us a little bit and we could still see that the underrepresented classes get mixed up with each other. \n",
        "\n",
        "Due to the great imbalance of the dataset, it is much better to re-consider our problem and instead of classifying signals which are hugely under-represented, we could try learning features from the strongly represented class and do an anomaly detection instead.\n",
        "\n",
        "\n",
        "\n",
        "---\n",
        "\n",
        "Interactive of our example: https://anomagram.fastforwardlabs.com/#/\n",
        "\n",
        "You can watch a tutorial here: https://www.youtube.com/watch?v=2K3ScZp1dXQ#/\n"
        
        
      ],
      "metadata": {
        "id": "gYJQ7eGcW6nS"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Few jupyter notebook shortcuts you can use to make your life easier**\n",
        "\n",
        "Run cell: `shift + enter`\n",
        "\n",
        "Undo last action (inside a cell): `ctrl + m + z`\n",
        "\n",
        "Find and replace: `ctrl + h`\n",
        "\n",
        "Insert code cell above: `ctrl + a`\n",
        "\n",
        "Insert code cell below: `ctrl + b`\n",
        "\n",
        "Delete cell: `ctrl + m + d`"
      ],
      "metadata": {
        "id": "nNMzXAoP2irN"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let's download the data"
      ],
      "metadata": {
        "id": "H5T2AO5cW6yP"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "amFW0AIpQpcs",
        "outputId": "0831a81a-10d1-4f43-ef14-e1d73c14cc5f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2021-12-18 14:03:39--  http://www.timeseriesclassification.com/Downloads/ECG5000.zip\n",
            "Resolving www.timeseriesclassification.com (www.timeseriesclassification.com)... 109.123.71.232\n",
            "Connecting to www.timeseriesclassification.com (www.timeseriesclassification.com)|109.123.71.232|:80... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 10614407 (10M) [application/zip]\n",
            "Saving to: ‘ECG5000.zip.1’\n",
            "\n",
            "ECG5000.zip.1       100%[===================>]  10.12M  4.11MB/s    in 2.5s    \n",
            "\n",
            "2021-12-18 14:03:42 (4.11 MB/s) - ‘ECG5000.zip.1’ saved [10614407/10614407]\n",
            "\n",
            "Archive:  ECG5000.zip\n",
            "replace ECG5000.txt? [y]es, [n]o, [A]ll, [N]one, [r]ename: N\n"
          ]
        }
      ],
      "source": [
        "! wget http://www.timeseriesclassification.com/Downloads/ECG5000.zip\n",
        "! unzip ECG5000.zip"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "And now since we don't need train and test files, we can concatenate the two files together and create one file containing all the data instead.\n",
        "\n",
        "Here we can simply do this by the the unix command `cat`"
      ],
      "metadata": {
        "id": "2-Nk8Re8W2DU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!cat ECG5000_TRAIN.txt ECG5000_TEST.txt > ecg_concat.txt"
      ],
      "metadata": {
        "id": "y8yU2IiTZNI1"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "And we can check if the file-structure still looks the same "
      ],
      "metadata": {
        "id": "3u4HCndxY9wv"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!head ecg_concat.txt"
      ],
      "metadata": {
        "id": "jZBEOotmZYiD"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let's import libraries we are about to use"
      ],
      "metadata": {
        "id": "mL5ge9IJZFdi"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras.models import Model\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn import preprocessing"
      ],
      "metadata": {
        "id": "wbWNtDuFYySV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Let's prepare our data"
      ],
      "metadata": {
        "id": "QcuQvU6fZN9e"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Once again, we create a dataframe with pandas..."
      ],
      "metadata": {
        "id": "IdRq3rb4ZU_V"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "df = pd.read_csv('/content/ecg_concat.txt', sep='  ', header=None)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DtqkASDDZBDB",
        "outputId": "948e6715-a555-4f9e-c0f2-453fa981bd6c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:1: ParserWarning: Falling back to the 'python' engine because the 'c' engine does not support regex separators (separators > 1 char and different from '\\s+' are interpreted as regex); you can avoid this warning by specifying engine='python'.\n",
            "  \"\"\"Entry point for launching an IPython kernel.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 488
        },
        "id": "szXAJk6AZicY",
        "outputId": "5ee0f363-f8d4-4233-da54-352fd12ff1b9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-b4a6ff79-7ca6-4867-bff3-6724d789a2d7\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "      <th>4</th>\n",
              "      <th>5</th>\n",
              "      <th>6</th>\n",
              "      <th>7</th>\n",
              "      <th>8</th>\n",
              "      <th>9</th>\n",
              "      <th>10</th>\n",
              "      <th>11</th>\n",
              "      <th>12</th>\n",
              "      <th>13</th>\n",
              "      <th>14</th>\n",
              "      <th>15</th>\n",
              "      <th>16</th>\n",
              "      <th>17</th>\n",
              "      <th>18</th>\n",
              "      <th>19</th>\n",
              "      <th>20</th>\n",
              "      <th>21</th>\n",
              "      <th>22</th>\n",
              "      <th>23</th>\n",
              "      <th>24</th>\n",
              "      <th>25</th>\n",
              "      <th>26</th>\n",
              "      <th>27</th>\n",
              "      <th>28</th>\n",
              "      <th>29</th>\n",
              "      <th>30</th>\n",
              "      <th>31</th>\n",
              "      <th>32</th>\n",
              "      <th>33</th>\n",
              "      <th>34</th>\n",
              "      <th>35</th>\n",
              "      <th>36</th>\n",
              "      <th>37</th>\n",
              "      <th>38</th>\n",
              "      <th>39</th>\n",
              "      <th>...</th>\n",
              "      <th>101</th>\n",
              "      <th>102</th>\n",
              "      <th>103</th>\n",
              "      <th>104</th>\n",
              "      <th>105</th>\n",
              "      <th>106</th>\n",
              "      <th>107</th>\n",
              "      <th>108</th>\n",
              "      <th>109</th>\n",
              "      <th>110</th>\n",
              "      <th>111</th>\n",
              "      <th>112</th>\n",
              "      <th>113</th>\n",
              "      <th>114</th>\n",
              "      <th>115</th>\n",
              "      <th>116</th>\n",
              "      <th>117</th>\n",
              "      <th>118</th>\n",
              "      <th>119</th>\n",
              "      <th>120</th>\n",
              "      <th>121</th>\n",
              "      <th>122</th>\n",
              "      <th>123</th>\n",
              "      <th>124</th>\n",
              "      <th>125</th>\n",
              "      <th>126</th>\n",
              "      <th>127</th>\n",
              "      <th>128</th>\n",
              "      <th>129</th>\n",
              "      <th>130</th>\n",
              "      <th>131</th>\n",
              "      <th>132</th>\n",
              "      <th>133</th>\n",
              "      <th>134</th>\n",
              "      <th>135</th>\n",
              "      <th>136</th>\n",
              "      <th>137</th>\n",
              "      <th>138</th>\n",
              "      <th>139</th>\n",
              "      <th>140</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1.0</td>\n",
              "      <td>-0.112522</td>\n",
              "      <td>-2.827204</td>\n",
              "      <td>-3.773897</td>\n",
              "      <td>-4.349751</td>\n",
              "      <td>-4.376041</td>\n",
              "      <td>-3.474986</td>\n",
              "      <td>-2.181408</td>\n",
              "      <td>-1.818287</td>\n",
              "      <td>-1.250522</td>\n",
              "      <td>-0.477492</td>\n",
              "      <td>-0.363808</td>\n",
              "      <td>-0.491957</td>\n",
              "      <td>-0.421855</td>\n",
              "      <td>-0.309201</td>\n",
              "      <td>-0.495939</td>\n",
              "      <td>-0.342119</td>\n",
              "      <td>-0.355336</td>\n",
              "      <td>-0.367913</td>\n",
              "      <td>-0.316503</td>\n",
              "      <td>-0.412374</td>\n",
              "      <td>-0.471672</td>\n",
              "      <td>-0.413458</td>\n",
              "      <td>-0.364617</td>\n",
              "      <td>-0.449298</td>\n",
              "      <td>-0.471419</td>\n",
              "      <td>-0.424777</td>\n",
              "      <td>-0.462517</td>\n",
              "      <td>-0.552472</td>\n",
              "      <td>-0.475375</td>\n",
              "      <td>-0.694200</td>\n",
              "      <td>-0.701868</td>\n",
              "      <td>-0.593812</td>\n",
              "      <td>-0.660684</td>\n",
              "      <td>-0.713831</td>\n",
              "      <td>-0.769807</td>\n",
              "      <td>-0.672282</td>\n",
              "      <td>-0.653676</td>\n",
              "      <td>-0.639406</td>\n",
              "      <td>-0.559302</td>\n",
              "      <td>...</td>\n",
              "      <td>1.173345</td>\n",
              "      <td>1.258179</td>\n",
              "      <td>1.433789</td>\n",
              "      <td>1.700533</td>\n",
              "      <td>1.999043</td>\n",
              "      <td>2.125341</td>\n",
              "      <td>1.993291</td>\n",
              "      <td>1.932246</td>\n",
              "      <td>1.797437</td>\n",
              "      <td>1.522284</td>\n",
              "      <td>1.251168</td>\n",
              "      <td>0.998730</td>\n",
              "      <td>0.483722</td>\n",
              "      <td>0.023132</td>\n",
              "      <td>-0.194914</td>\n",
              "      <td>-0.220917</td>\n",
              "      <td>-0.243737</td>\n",
              "      <td>-0.254695</td>\n",
              "      <td>-0.291136</td>\n",
              "      <td>-0.256490</td>\n",
              "      <td>-0.227874</td>\n",
              "      <td>-0.322423</td>\n",
              "      <td>-0.289286</td>\n",
              "      <td>-0.318170</td>\n",
              "      <td>-0.363654</td>\n",
              "      <td>-0.393456</td>\n",
              "      <td>-0.266419</td>\n",
              "      <td>-0.256823</td>\n",
              "      <td>-0.288694</td>\n",
              "      <td>-0.162338</td>\n",
              "      <td>0.160348</td>\n",
              "      <td>0.792168</td>\n",
              "      <td>0.933541</td>\n",
              "      <td>0.796958</td>\n",
              "      <td>0.578621</td>\n",
              "      <td>0.257740</td>\n",
              "      <td>0.228077</td>\n",
              "      <td>0.123431</td>\n",
              "      <td>0.925286</td>\n",
              "      <td>0.193137</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1.0</td>\n",
              "      <td>-1.100878</td>\n",
              "      <td>-3.996840</td>\n",
              "      <td>-4.285843</td>\n",
              "      <td>-4.506579</td>\n",
              "      <td>-4.022377</td>\n",
              "      <td>-3.234368</td>\n",
              "      <td>-1.566126</td>\n",
              "      <td>-0.992258</td>\n",
              "      <td>-0.754680</td>\n",
              "      <td>0.042321</td>\n",
              "      <td>0.148951</td>\n",
              "      <td>0.183527</td>\n",
              "      <td>0.294876</td>\n",
              "      <td>0.190233</td>\n",
              "      <td>0.235575</td>\n",
              "      <td>0.253487</td>\n",
              "      <td>0.221742</td>\n",
              "      <td>0.050233</td>\n",
              "      <td>0.178042</td>\n",
              "      <td>0.139563</td>\n",
              "      <td>0.046794</td>\n",
              "      <td>0.043007</td>\n",
              "      <td>0.106544</td>\n",
              "      <td>0.012654</td>\n",
              "      <td>0.003995</td>\n",
              "      <td>0.045724</td>\n",
              "      <td>-0.045999</td>\n",
              "      <td>-0.072667</td>\n",
              "      <td>-0.071078</td>\n",
              "      <td>-0.153866</td>\n",
              "      <td>-0.227254</td>\n",
              "      <td>-0.249270</td>\n",
              "      <td>-0.253489</td>\n",
              "      <td>-0.332835</td>\n",
              "      <td>-0.264330</td>\n",
              "      <td>-0.345825</td>\n",
              "      <td>-0.310781</td>\n",
              "      <td>-0.334160</td>\n",
              "      <td>-0.306178</td>\n",
              "      <td>...</td>\n",
              "      <td>1.557410</td>\n",
              "      <td>1.808428</td>\n",
              "      <td>2.164346</td>\n",
              "      <td>2.070747</td>\n",
              "      <td>1.903614</td>\n",
              "      <td>1.764455</td>\n",
              "      <td>1.507769</td>\n",
              "      <td>1.293428</td>\n",
              "      <td>0.894562</td>\n",
              "      <td>0.578016</td>\n",
              "      <td>0.244343</td>\n",
              "      <td>-0.286443</td>\n",
              "      <td>-0.515881</td>\n",
              "      <td>-0.732707</td>\n",
              "      <td>-0.832465</td>\n",
              "      <td>-0.803318</td>\n",
              "      <td>-0.836252</td>\n",
              "      <td>-0.777865</td>\n",
              "      <td>-0.774753</td>\n",
              "      <td>-0.733404</td>\n",
              "      <td>-0.721386</td>\n",
              "      <td>-0.832095</td>\n",
              "      <td>-0.711982</td>\n",
              "      <td>-0.751867</td>\n",
              "      <td>-0.757720</td>\n",
              "      <td>-0.853120</td>\n",
              "      <td>-0.766988</td>\n",
              "      <td>-0.688161</td>\n",
              "      <td>-0.519923</td>\n",
              "      <td>0.039406</td>\n",
              "      <td>0.560327</td>\n",
              "      <td>0.538356</td>\n",
              "      <td>0.656881</td>\n",
              "      <td>0.787490</td>\n",
              "      <td>0.724046</td>\n",
              "      <td>0.555784</td>\n",
              "      <td>0.476333</td>\n",
              "      <td>0.773820</td>\n",
              "      <td>1.119621</td>\n",
              "      <td>-1.436250</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1.0</td>\n",
              "      <td>-0.567088</td>\n",
              "      <td>-2.593450</td>\n",
              "      <td>-3.874230</td>\n",
              "      <td>-4.584095</td>\n",
              "      <td>-4.187449</td>\n",
              "      <td>-3.151462</td>\n",
              "      <td>-1.742940</td>\n",
              "      <td>-1.490658</td>\n",
              "      <td>-1.183580</td>\n",
              "      <td>-0.394229</td>\n",
              "      <td>-0.282897</td>\n",
              "      <td>-0.356926</td>\n",
              "      <td>-0.287297</td>\n",
              "      <td>-0.399489</td>\n",
              "      <td>-0.473244</td>\n",
              "      <td>-0.379048</td>\n",
              "      <td>-0.399039</td>\n",
              "      <td>-0.178594</td>\n",
              "      <td>-0.339522</td>\n",
              "      <td>-0.498447</td>\n",
              "      <td>-0.337251</td>\n",
              "      <td>-0.425480</td>\n",
              "      <td>-0.423952</td>\n",
              "      <td>-0.463170</td>\n",
              "      <td>-0.493253</td>\n",
              "      <td>-0.549749</td>\n",
              "      <td>-0.529831</td>\n",
              "      <td>-0.530935</td>\n",
              "      <td>-0.502365</td>\n",
              "      <td>-0.417368</td>\n",
              "      <td>-0.526346</td>\n",
              "      <td>-0.471005</td>\n",
              "      <td>-0.676784</td>\n",
              "      <td>-0.898612</td>\n",
              "      <td>-0.610571</td>\n",
              "      <td>-0.530164</td>\n",
              "      <td>-0.765674</td>\n",
              "      <td>-0.581937</td>\n",
              "      <td>-0.537848</td>\n",
              "      <td>...</td>\n",
              "      <td>1.640105</td>\n",
              "      <td>1.810988</td>\n",
              "      <td>2.185398</td>\n",
              "      <td>2.262985</td>\n",
              "      <td>2.052920</td>\n",
              "      <td>1.890488</td>\n",
              "      <td>1.793033</td>\n",
              "      <td>1.564784</td>\n",
              "      <td>1.234619</td>\n",
              "      <td>0.900302</td>\n",
              "      <td>0.551957</td>\n",
              "      <td>0.258222</td>\n",
              "      <td>-0.128587</td>\n",
              "      <td>-0.092585</td>\n",
              "      <td>-0.168606</td>\n",
              "      <td>-0.495989</td>\n",
              "      <td>-0.395034</td>\n",
              "      <td>-0.328238</td>\n",
              "      <td>-0.448138</td>\n",
              "      <td>-0.268230</td>\n",
              "      <td>-0.456415</td>\n",
              "      <td>-0.357867</td>\n",
              "      <td>-0.317508</td>\n",
              "      <td>-0.434112</td>\n",
              "      <td>-0.549203</td>\n",
              "      <td>-0.324615</td>\n",
              "      <td>-0.268082</td>\n",
              "      <td>-0.220384</td>\n",
              "      <td>-0.117429</td>\n",
              "      <td>0.614059</td>\n",
              "      <td>1.284825</td>\n",
              "      <td>0.886073</td>\n",
              "      <td>0.531452</td>\n",
              "      <td>0.311377</td>\n",
              "      <td>-0.021919</td>\n",
              "      <td>-0.713683</td>\n",
              "      <td>-0.532197</td>\n",
              "      <td>0.321097</td>\n",
              "      <td>0.904227</td>\n",
              "      <td>-0.421797</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.490473</td>\n",
              "      <td>-1.914407</td>\n",
              "      <td>-3.616364</td>\n",
              "      <td>-4.318823</td>\n",
              "      <td>-4.268016</td>\n",
              "      <td>-3.881110</td>\n",
              "      <td>-2.993280</td>\n",
              "      <td>-1.671131</td>\n",
              "      <td>-1.333884</td>\n",
              "      <td>-0.965629</td>\n",
              "      <td>-0.183319</td>\n",
              "      <td>-0.101657</td>\n",
              "      <td>-0.273874</td>\n",
              "      <td>-0.127818</td>\n",
              "      <td>-0.195983</td>\n",
              "      <td>-0.213523</td>\n",
              "      <td>-0.176473</td>\n",
              "      <td>-0.156932</td>\n",
              "      <td>-0.149172</td>\n",
              "      <td>-0.181510</td>\n",
              "      <td>-0.180074</td>\n",
              "      <td>-0.246151</td>\n",
              "      <td>-0.274260</td>\n",
              "      <td>-0.140960</td>\n",
              "      <td>-0.277449</td>\n",
              "      <td>-0.382549</td>\n",
              "      <td>-0.311937</td>\n",
              "      <td>-0.360093</td>\n",
              "      <td>-0.405968</td>\n",
              "      <td>-0.571433</td>\n",
              "      <td>-0.524106</td>\n",
              "      <td>-0.537886</td>\n",
              "      <td>-0.606778</td>\n",
              "      <td>-0.661446</td>\n",
              "      <td>-0.683375</td>\n",
              "      <td>-0.746683</td>\n",
              "      <td>-0.635662</td>\n",
              "      <td>-0.625231</td>\n",
              "      <td>-0.540094</td>\n",
              "      <td>...</td>\n",
              "      <td>1.553079</td>\n",
              "      <td>1.772155</td>\n",
              "      <td>2.000769</td>\n",
              "      <td>1.925003</td>\n",
              "      <td>1.898426</td>\n",
              "      <td>1.720953</td>\n",
              "      <td>1.501711</td>\n",
              "      <td>1.422492</td>\n",
              "      <td>1.023225</td>\n",
              "      <td>0.776341</td>\n",
              "      <td>0.504426</td>\n",
              "      <td>0.056382</td>\n",
              "      <td>-0.233161</td>\n",
              "      <td>-0.406388</td>\n",
              "      <td>-0.327528</td>\n",
              "      <td>-0.460868</td>\n",
              "      <td>-0.402536</td>\n",
              "      <td>-0.345752</td>\n",
              "      <td>-0.354206</td>\n",
              "      <td>-0.439959</td>\n",
              "      <td>-0.425326</td>\n",
              "      <td>-0.439789</td>\n",
              "      <td>-0.451835</td>\n",
              "      <td>-0.395926</td>\n",
              "      <td>-0.448762</td>\n",
              "      <td>-0.391789</td>\n",
              "      <td>-0.376307</td>\n",
              "      <td>-0.461069</td>\n",
              "      <td>-0.253524</td>\n",
              "      <td>0.213006</td>\n",
              "      <td>0.491173</td>\n",
              "      <td>0.350816</td>\n",
              "      <td>0.499111</td>\n",
              "      <td>0.600345</td>\n",
              "      <td>0.842069</td>\n",
              "      <td>0.952074</td>\n",
              "      <td>0.990133</td>\n",
              "      <td>1.086798</td>\n",
              "      <td>1.403011</td>\n",
              "      <td>-0.383564</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.800232</td>\n",
              "      <td>-0.874252</td>\n",
              "      <td>-2.384761</td>\n",
              "      <td>-3.973292</td>\n",
              "      <td>-4.338224</td>\n",
              "      <td>-3.802422</td>\n",
              "      <td>-2.534510</td>\n",
              "      <td>-1.783423</td>\n",
              "      <td>-1.594450</td>\n",
              "      <td>-0.753199</td>\n",
              "      <td>-0.298107</td>\n",
              "      <td>-0.428928</td>\n",
              "      <td>-0.491351</td>\n",
              "      <td>-0.361304</td>\n",
              "      <td>-0.339296</td>\n",
              "      <td>-0.324952</td>\n",
              "      <td>-0.290113</td>\n",
              "      <td>-0.363051</td>\n",
              "      <td>-0.525684</td>\n",
              "      <td>-0.597423</td>\n",
              "      <td>-0.575523</td>\n",
              "      <td>-0.567503</td>\n",
              "      <td>-0.504555</td>\n",
              "      <td>-0.618406</td>\n",
              "      <td>-0.682814</td>\n",
              "      <td>-0.743849</td>\n",
              "      <td>-0.815588</td>\n",
              "      <td>-0.826902</td>\n",
              "      <td>-0.782374</td>\n",
              "      <td>-0.929462</td>\n",
              "      <td>-0.999672</td>\n",
              "      <td>-1.060969</td>\n",
              "      <td>-1.007877</td>\n",
              "      <td>-1.028735</td>\n",
              "      <td>-1.122629</td>\n",
              "      <td>-1.028650</td>\n",
              "      <td>-1.046515</td>\n",
              "      <td>-1.063372</td>\n",
              "      <td>-1.122423</td>\n",
              "      <td>...</td>\n",
              "      <td>0.982230</td>\n",
              "      <td>1.155363</td>\n",
              "      <td>1.336254</td>\n",
              "      <td>1.627534</td>\n",
              "      <td>1.717594</td>\n",
              "      <td>1.696487</td>\n",
              "      <td>1.741686</td>\n",
              "      <td>1.674078</td>\n",
              "      <td>1.546928</td>\n",
              "      <td>1.331738</td>\n",
              "      <td>1.110168</td>\n",
              "      <td>0.922210</td>\n",
              "      <td>0.521777</td>\n",
              "      <td>0.154852</td>\n",
              "      <td>-0.123861</td>\n",
              "      <td>-0.202998</td>\n",
              "      <td>-0.247956</td>\n",
              "      <td>-0.219122</td>\n",
              "      <td>-0.214695</td>\n",
              "      <td>-0.319215</td>\n",
              "      <td>-0.198597</td>\n",
              "      <td>-0.151618</td>\n",
              "      <td>-0.129593</td>\n",
              "      <td>-0.074939</td>\n",
              "      <td>-0.196807</td>\n",
              "      <td>-0.174795</td>\n",
              "      <td>-0.208833</td>\n",
              "      <td>-0.210754</td>\n",
              "      <td>-0.100485</td>\n",
              "      <td>0.197446</td>\n",
              "      <td>0.966606</td>\n",
              "      <td>1.148884</td>\n",
              "      <td>0.958434</td>\n",
              "      <td>1.059025</td>\n",
              "      <td>1.371682</td>\n",
              "      <td>1.277392</td>\n",
              "      <td>0.960304</td>\n",
              "      <td>0.971020</td>\n",
              "      <td>1.614392</td>\n",
              "      <td>1.421456</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4995</th>\n",
              "      <td>4.0</td>\n",
              "      <td>-1.122969</td>\n",
              "      <td>-2.252925</td>\n",
              "      <td>-2.867628</td>\n",
              "      <td>-3.358605</td>\n",
              "      <td>-3.167849</td>\n",
              "      <td>-2.638360</td>\n",
              "      <td>-1.664162</td>\n",
              "      <td>-0.935655</td>\n",
              "      <td>-0.866953</td>\n",
              "      <td>-0.645363</td>\n",
              "      <td>-0.093342</td>\n",
              "      <td>0.099746</td>\n",
              "      <td>0.121585</td>\n",
              "      <td>0.125760</td>\n",
              "      <td>0.071543</td>\n",
              "      <td>0.050503</td>\n",
              "      <td>0.021200</td>\n",
              "      <td>0.026780</td>\n",
              "      <td>0.158262</td>\n",
              "      <td>0.099955</td>\n",
              "      <td>-0.009558</td>\n",
              "      <td>0.078789</td>\n",
              "      <td>0.086691</td>\n",
              "      <td>0.046382</td>\n",
              "      <td>0.068272</td>\n",
              "      <td>0.102911</td>\n",
              "      <td>0.106083</td>\n",
              "      <td>0.057099</td>\n",
              "      <td>0.028555</td>\n",
              "      <td>0.046705</td>\n",
              "      <td>0.046524</td>\n",
              "      <td>0.095233</td>\n",
              "      <td>0.059824</td>\n",
              "      <td>-0.021798</td>\n",
              "      <td>-0.022737</td>\n",
              "      <td>-0.083263</td>\n",
              "      <td>-0.125486</td>\n",
              "      <td>-0.082084</td>\n",
              "      <td>-0.023000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.643776</td>\n",
              "      <td>0.597168</td>\n",
              "      <td>0.625141</td>\n",
              "      <td>0.648181</td>\n",
              "      <td>0.594234</td>\n",
              "      <td>0.630589</td>\n",
              "      <td>0.654510</td>\n",
              "      <td>0.602251</td>\n",
              "      <td>0.656946</td>\n",
              "      <td>0.603251</td>\n",
              "      <td>0.599274</td>\n",
              "      <td>0.561487</td>\n",
              "      <td>0.547185</td>\n",
              "      <td>0.537524</td>\n",
              "      <td>0.603885</td>\n",
              "      <td>0.541677</td>\n",
              "      <td>0.580715</td>\n",
              "      <td>0.689346</td>\n",
              "      <td>0.746181</td>\n",
              "      <td>0.867390</td>\n",
              "      <td>0.783529</td>\n",
              "      <td>0.698720</td>\n",
              "      <td>0.798727</td>\n",
              "      <td>0.875711</td>\n",
              "      <td>0.985118</td>\n",
              "      <td>0.799260</td>\n",
              "      <td>0.577579</td>\n",
              "      <td>0.394182</td>\n",
              "      <td>0.458721</td>\n",
              "      <td>0.411458</td>\n",
              "      <td>0.205543</td>\n",
              "      <td>-0.472419</td>\n",
              "      <td>-1.310147</td>\n",
              "      <td>-2.029521</td>\n",
              "      <td>-3.221294</td>\n",
              "      <td>-4.176790</td>\n",
              "      <td>-4.009720</td>\n",
              "      <td>-2.874136</td>\n",
              "      <td>-2.008369</td>\n",
              "      <td>-1.808334</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4996</th>\n",
              "      <td>2.0</td>\n",
              "      <td>-0.547705</td>\n",
              "      <td>-1.889545</td>\n",
              "      <td>-2.839779</td>\n",
              "      <td>-3.457912</td>\n",
              "      <td>-3.929149</td>\n",
              "      <td>-3.966026</td>\n",
              "      <td>-3.492560</td>\n",
              "      <td>-2.695269</td>\n",
              "      <td>-1.849691</td>\n",
              "      <td>-1.374321</td>\n",
              "      <td>-1.288459</td>\n",
              "      <td>-1.117578</td>\n",
              "      <td>-0.536271</td>\n",
              "      <td>-0.251050</td>\n",
              "      <td>-0.215496</td>\n",
              "      <td>-0.072500</td>\n",
              "      <td>-0.170556</td>\n",
              "      <td>-0.182279</td>\n",
              "      <td>-0.271306</td>\n",
              "      <td>-0.242693</td>\n",
              "      <td>-0.225397</td>\n",
              "      <td>-0.187956</td>\n",
              "      <td>-0.215918</td>\n",
              "      <td>-0.173543</td>\n",
              "      <td>-0.233401</td>\n",
              "      <td>-0.190231</td>\n",
              "      <td>-0.160668</td>\n",
              "      <td>-0.241456</td>\n",
              "      <td>-0.204146</td>\n",
              "      <td>-0.228155</td>\n",
              "      <td>-0.084758</td>\n",
              "      <td>-0.142782</td>\n",
              "      <td>-0.153214</td>\n",
              "      <td>-0.314872</td>\n",
              "      <td>-0.335928</td>\n",
              "      <td>-0.266875</td>\n",
              "      <td>-0.313519</td>\n",
              "      <td>-0.304439</td>\n",
              "      <td>-0.294795</td>\n",
              "      <td>...</td>\n",
              "      <td>0.427870</td>\n",
              "      <td>0.471064</td>\n",
              "      <td>0.532288</td>\n",
              "      <td>0.541352</td>\n",
              "      <td>0.588983</td>\n",
              "      <td>0.753851</td>\n",
              "      <td>0.650792</td>\n",
              "      <td>0.720441</td>\n",
              "      <td>0.752308</td>\n",
              "      <td>0.813847</td>\n",
              "      <td>0.690120</td>\n",
              "      <td>0.685244</td>\n",
              "      <td>0.632990</td>\n",
              "      <td>0.688632</td>\n",
              "      <td>0.676820</td>\n",
              "      <td>0.639733</td>\n",
              "      <td>0.708132</td>\n",
              "      <td>0.675063</td>\n",
              "      <td>0.646640</td>\n",
              "      <td>0.635341</td>\n",
              "      <td>0.670459</td>\n",
              "      <td>0.732434</td>\n",
              "      <td>0.866642</td>\n",
              "      <td>0.845284</td>\n",
              "      <td>1.045767</td>\n",
              "      <td>1.429693</td>\n",
              "      <td>1.783791</td>\n",
              "      <td>1.697750</td>\n",
              "      <td>1.492639</td>\n",
              "      <td>1.378843</td>\n",
              "      <td>1.218185</td>\n",
              "      <td>1.258419</td>\n",
              "      <td>1.907529</td>\n",
              "      <td>2.280888</td>\n",
              "      <td>1.895242</td>\n",
              "      <td>1.437702</td>\n",
              "      <td>1.193433</td>\n",
              "      <td>1.261335</td>\n",
              "      <td>1.150449</td>\n",
              "      <td>0.804932</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4997</th>\n",
              "      <td>2.0</td>\n",
              "      <td>-1.351779</td>\n",
              "      <td>-2.209006</td>\n",
              "      <td>-2.520225</td>\n",
              "      <td>-3.061475</td>\n",
              "      <td>-3.065141</td>\n",
              "      <td>-3.030739</td>\n",
              "      <td>-2.622720</td>\n",
              "      <td>-2.044092</td>\n",
              "      <td>-1.295874</td>\n",
              "      <td>-0.733839</td>\n",
              "      <td>-0.562769</td>\n",
              "      <td>-0.461068</td>\n",
              "      <td>-0.310064</td>\n",
              "      <td>-0.004490</td>\n",
              "      <td>0.184161</td>\n",
              "      <td>0.252896</td>\n",
              "      <td>0.302077</td>\n",
              "      <td>0.237223</td>\n",
              "      <td>0.304171</td>\n",
              "      <td>0.276120</td>\n",
              "      <td>0.330877</td>\n",
              "      <td>0.356674</td>\n",
              "      <td>0.323301</td>\n",
              "      <td>0.413409</td>\n",
              "      <td>0.366606</td>\n",
              "      <td>0.328516</td>\n",
              "      <td>0.358460</td>\n",
              "      <td>0.353359</td>\n",
              "      <td>0.451590</td>\n",
              "      <td>0.419947</td>\n",
              "      <td>0.330580</td>\n",
              "      <td>0.317684</td>\n",
              "      <td>0.270813</td>\n",
              "      <td>0.258002</td>\n",
              "      <td>0.317464</td>\n",
              "      <td>0.358579</td>\n",
              "      <td>0.295808</td>\n",
              "      <td>0.276734</td>\n",
              "      <td>0.230999</td>\n",
              "      <td>...</td>\n",
              "      <td>0.716374</td>\n",
              "      <td>0.802243</td>\n",
              "      <td>0.803918</td>\n",
              "      <td>0.826138</td>\n",
              "      <td>0.774642</td>\n",
              "      <td>0.637822</td>\n",
              "      <td>0.747995</td>\n",
              "      <td>0.762681</td>\n",
              "      <td>0.807652</td>\n",
              "      <td>0.816406</td>\n",
              "      <td>0.818481</td>\n",
              "      <td>0.900709</td>\n",
              "      <td>0.876553</td>\n",
              "      <td>0.861227</td>\n",
              "      <td>0.764280</td>\n",
              "      <td>0.628381</td>\n",
              "      <td>0.485624</td>\n",
              "      <td>0.462129</td>\n",
              "      <td>0.442520</td>\n",
              "      <td>0.402201</td>\n",
              "      <td>0.380095</td>\n",
              "      <td>0.289096</td>\n",
              "      <td>0.214385</td>\n",
              "      <td>0.106227</td>\n",
              "      <td>-0.009595</td>\n",
              "      <td>-0.025391</td>\n",
              "      <td>0.035677</td>\n",
              "      <td>-0.030661</td>\n",
              "      <td>-0.203568</td>\n",
              "      <td>-0.445023</td>\n",
              "      <td>-0.896575</td>\n",
              "      <td>-1.512234</td>\n",
              "      <td>-2.076075</td>\n",
              "      <td>-2.586042</td>\n",
              "      <td>-3.322799</td>\n",
              "      <td>-3.627311</td>\n",
              "      <td>-3.437038</td>\n",
              "      <td>-2.260023</td>\n",
              "      <td>-1.577823</td>\n",
              "      <td>-0.684531</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4998</th>\n",
              "      <td>2.0</td>\n",
              "      <td>-1.124432</td>\n",
              "      <td>-1.905039</td>\n",
              "      <td>-2.192707</td>\n",
              "      <td>-2.904320</td>\n",
              "      <td>-2.900722</td>\n",
              "      <td>-2.761252</td>\n",
              "      <td>-2.569705</td>\n",
              "      <td>-2.043893</td>\n",
              "      <td>-1.490538</td>\n",
              "      <td>-0.938473</td>\n",
              "      <td>-0.485372</td>\n",
              "      <td>-0.374298</td>\n",
              "      <td>-0.389414</td>\n",
              "      <td>-0.146131</td>\n",
              "      <td>0.212746</td>\n",
              "      <td>0.429797</td>\n",
              "      <td>0.499480</td>\n",
              "      <td>0.536705</td>\n",
              "      <td>0.558162</td>\n",
              "      <td>0.543721</td>\n",
              "      <td>0.550670</td>\n",
              "      <td>0.560516</td>\n",
              "      <td>0.537170</td>\n",
              "      <td>0.537986</td>\n",
              "      <td>0.549989</td>\n",
              "      <td>0.519434</td>\n",
              "      <td>0.508304</td>\n",
              "      <td>0.540967</td>\n",
              "      <td>0.570611</td>\n",
              "      <td>0.587472</td>\n",
              "      <td>0.556779</td>\n",
              "      <td>0.519021</td>\n",
              "      <td>0.507215</td>\n",
              "      <td>0.464078</td>\n",
              "      <td>0.450806</td>\n",
              "      <td>0.475988</td>\n",
              "      <td>0.485856</td>\n",
              "      <td>0.466050</td>\n",
              "      <td>0.392451</td>\n",
              "      <td>...</td>\n",
              "      <td>0.762903</td>\n",
              "      <td>0.757111</td>\n",
              "      <td>0.704867</td>\n",
              "      <td>0.726405</td>\n",
              "      <td>0.720765</td>\n",
              "      <td>0.769695</td>\n",
              "      <td>0.809712</td>\n",
              "      <td>0.703485</td>\n",
              "      <td>0.610350</td>\n",
              "      <td>0.683131</td>\n",
              "      <td>0.746369</td>\n",
              "      <td>0.605945</td>\n",
              "      <td>0.437233</td>\n",
              "      <td>0.395451</td>\n",
              "      <td>0.396267</td>\n",
              "      <td>0.396361</td>\n",
              "      <td>0.354488</td>\n",
              "      <td>0.279633</td>\n",
              "      <td>0.227358</td>\n",
              "      <td>0.123992</td>\n",
              "      <td>-0.031222</td>\n",
              "      <td>-0.155341</td>\n",
              "      <td>-0.219141</td>\n",
              "      <td>-0.239644</td>\n",
              "      <td>-0.333443</td>\n",
              "      <td>-0.557080</td>\n",
              "      <td>-0.811259</td>\n",
              "      <td>-1.072168</td>\n",
              "      <td>-1.497530</td>\n",
              "      <td>-2.047796</td>\n",
              "      <td>-2.495989</td>\n",
              "      <td>-2.821782</td>\n",
              "      <td>-3.268355</td>\n",
              "      <td>-3.634981</td>\n",
              "      <td>-3.168765</td>\n",
              "      <td>-2.245878</td>\n",
              "      <td>-1.262259</td>\n",
              "      <td>-0.443307</td>\n",
              "      <td>-0.559769</td>\n",
              "      <td>0.108568</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4999</th>\n",
              "      <td>2.0</td>\n",
              "      <td>0.728813</td>\n",
              "      <td>0.192597</td>\n",
              "      <td>-0.733884</td>\n",
              "      <td>-1.779456</td>\n",
              "      <td>-2.345908</td>\n",
              "      <td>-2.977565</td>\n",
              "      <td>-3.380053</td>\n",
              "      <td>-3.417164</td>\n",
              "      <td>-3.030925</td>\n",
              "      <td>-2.313867</td>\n",
              "      <td>-1.671269</td>\n",
              "      <td>-1.255303</td>\n",
              "      <td>-1.146982</td>\n",
              "      <td>-1.049088</td>\n",
              "      <td>-0.513524</td>\n",
              "      <td>-0.100892</td>\n",
              "      <td>-0.087888</td>\n",
              "      <td>-0.179308</td>\n",
              "      <td>-0.226885</td>\n",
              "      <td>-0.156335</td>\n",
              "      <td>-0.141478</td>\n",
              "      <td>-0.024932</td>\n",
              "      <td>-0.089615</td>\n",
              "      <td>-0.165480</td>\n",
              "      <td>-0.043243</td>\n",
              "      <td>-0.132201</td>\n",
              "      <td>-0.193009</td>\n",
              "      <td>-0.177407</td>\n",
              "      <td>-0.276008</td>\n",
              "      <td>-0.202072</td>\n",
              "      <td>-0.110621</td>\n",
              "      <td>-0.176676</td>\n",
              "      <td>-0.177095</td>\n",
              "      <td>-0.277154</td>\n",
              "      <td>-0.054093</td>\n",
              "      <td>-0.188261</td>\n",
              "      <td>-0.283490</td>\n",
              "      <td>-0.250398</td>\n",
              "      <td>-0.280567</td>\n",
              "      <td>...</td>\n",
              "      <td>0.343283</td>\n",
              "      <td>0.351328</td>\n",
              "      <td>0.350081</td>\n",
              "      <td>0.487850</td>\n",
              "      <td>0.514130</td>\n",
              "      <td>0.655713</td>\n",
              "      <td>0.659875</td>\n",
              "      <td>0.574770</td>\n",
              "      <td>0.675136</td>\n",
              "      <td>0.624451</td>\n",
              "      <td>0.697013</td>\n",
              "      <td>0.749857</td>\n",
              "      <td>0.829295</td>\n",
              "      <td>0.871171</td>\n",
              "      <td>0.863445</td>\n",
              "      <td>0.916639</td>\n",
              "      <td>0.841814</td>\n",
              "      <td>0.961598</td>\n",
              "      <td>0.877710</td>\n",
              "      <td>0.831334</td>\n",
              "      <td>0.774570</td>\n",
              "      <td>0.894903</td>\n",
              "      <td>1.087536</td>\n",
              "      <td>1.021885</td>\n",
              "      <td>1.208764</td>\n",
              "      <td>1.623936</td>\n",
              "      <td>1.851147</td>\n",
              "      <td>1.818792</td>\n",
              "      <td>1.672569</td>\n",
              "      <td>1.389086</td>\n",
              "      <td>1.219499</td>\n",
              "      <td>1.267275</td>\n",
              "      <td>1.678989</td>\n",
              "      <td>2.483389</td>\n",
              "      <td>2.569073</td>\n",
              "      <td>2.122891</td>\n",
              "      <td>1.753963</td>\n",
              "      <td>1.538975</td>\n",
              "      <td>1.713781</td>\n",
              "      <td>1.309382</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5000 rows × 141 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-b4a6ff79-7ca6-4867-bff3-6724d789a2d7')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-b4a6ff79-7ca6-4867-bff3-6724d789a2d7 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-b4a6ff79-7ca6-4867-bff3-6724d789a2d7');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ],
            "text/plain": [
              "      0         1         2         3    ...       137       138       139       140\n",
              "0     1.0 -0.112522 -2.827204 -3.773897  ...  0.228077  0.123431  0.925286  0.193137\n",
              "1     1.0 -1.100878 -3.996840 -4.285843  ...  0.476333  0.773820  1.119621 -1.436250\n",
              "2     1.0 -0.567088 -2.593450 -3.874230  ... -0.532197  0.321097  0.904227 -0.421797\n",
              "3     1.0  0.490473 -1.914407 -3.616364  ...  0.990133  1.086798  1.403011 -0.383564\n",
              "4     1.0  0.800232 -0.874252 -2.384761  ...  0.960304  0.971020  1.614392  1.421456\n",
              "...   ...       ...       ...       ...  ...       ...       ...       ...       ...\n",
              "4995  4.0 -1.122969 -2.252925 -2.867628  ... -4.009720 -2.874136 -2.008369 -1.808334\n",
              "4996  2.0 -0.547705 -1.889545 -2.839779  ...  1.193433  1.261335  1.150449  0.804932\n",
              "4997  2.0 -1.351779 -2.209006 -2.520225  ... -3.437038 -2.260023 -1.577823 -0.684531\n",
              "4998  2.0 -1.124432 -1.905039 -2.192707  ... -1.262259 -0.443307 -0.559769  0.108568\n",
              "4999  2.0  0.728813  0.192597 -0.733884  ...  1.753963  1.538975  1.713781  1.309382\n",
              "\n",
              "[5000 rows x 141 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "And add prefixes, cause by now we know we cannot work with the column names if they are just numerical..."
      ],
      "metadata": {
        "id": "RCsR8bxVaqOY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "df = df.add_prefix('c')\n",
        "df.columns"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JXwMvSI2ahgE",
        "outputId": "d48fd37f-8125-489e-bdda-ad63d655dc91"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['c0', 'c1', 'c2', 'c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9',\n",
              "       ...\n",
              "       'c131', 'c132', 'c133', 'c134', 'c135', 'c136', 'c137', 'c138', 'c139',\n",
              "       'c140'],\n",
              "      dtype='object', length=141)"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "So how many samples in each class do we have now? Let's see..."
      ],
      "metadata": {
        "id": "KSz0yBYXbCIg"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "df['c0'].value_counts()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rJ-UwBzOa7OZ",
        "outputId": "1da070e6-3c0e-4789-a730-0e4fb63eb07e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1.0    2919\n",
              "2.0    1767\n",
              "4.0     194\n",
              "3.0      96\n",
              "5.0      24\n",
              "Name: c0, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### We split the original data into train and test sets"
      ],
      "metadata": {
        "id": "u2YW5IdwZ9HK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x_train, x_test, y_train, y_test = train_test_split(df.values,\n",
        "                                                    df.values[:,0], \n",
        "                                                    test_size=0.2,\n",
        "                                                    random_state=42)  # by setting random state we will get the same data when recalling the function"
      ],
      "metadata": {
        "id": "KiBoCZoybP7V"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Now we normalize the data"
      ],
      "metadata": {
        "id": "1L3WKo7QoaL7"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We use MinMaxScaler that was import from sklearn\n",
        "\n",
        "Here the scaler holds the MinMaxScaler and we fit it on our training data in the second row of the code, which will be applied on the data we want to scale just in a second with `data_scaler.transform(data_we_want_to_scale)`"
      ],
      "metadata": {
        "id": "Wa4BIoHR0QLE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "scaler = preprocessing.MinMaxScaler()\n",
        "data_scaler = scaler.fit(x_train[:, 1:])"
      ],
      "metadata": {
        "id": "sGLySYu4wgBG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "We take columns 1 to 140 since the 0th column holds the classes and we don't want to change their value."
      ],
      "metadata": {
        "id": "zQS4nSGYDHeE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x_train[:, 1:] = data_scaler.transform(x_train[:, 1:])\n",
        "x_test[:, 1:] = data_scaler.transform(x_test[:, 1:])"
      ],
      "metadata": {
        "id": "fYO3rBCtDJU9"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Separate the data into normal and abnormal sets\n",
        "\n",
        "Well we know we have 5 classes, the first one with index 1 represents normal signals, whilst other 4 represents heart abnormalities.\n",
        "\n",
        "What we want is to have a class 0 that would represent the normal ECG signals and class 1 that holds all the abnormalities. Once again, pandas functions will help us with this. \n",
        "\n",
        "We also don't need the classes - the 0th column, so we take only the columns 1 to 140."
      ],
      "metadata": {
        "id": "e8CiWFsoceDP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "normal_x_train = pd.DataFrame(x_train).add_prefix('c').query('c0 == 1').values[:,1:]\n",
        "normal_x_test = pd.DataFrame(x_test).add_prefix('c').query('c0 == 1').values[:,1:]\n",
        "\n",
        "abnormal_x_train = pd.DataFrame(x_train).add_prefix('c').query('c0 > 1').values[:,1:]\n",
        "abnormal_x_test = pd.DataFrame(x_test).add_prefix('c').query('c0 > 1').values[:,1:]"
      ],
      "metadata": {
        "id": "ssu-gVHtBc07"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "**If you are fond of dataframes you could do this instead:**\n",
        "\n",
        "\n",
        "```\n",
        "x_train, x_test, y_train, y_test = train_test_split(df.values,\n",
        "                                                    df.values[:,0], \n",
        "                                                    test_size=0.2,\n",
        "                                                    random_state=42)  # by setting random state we will get the same data when recalling the function\n",
        "```\n",
        "\n",
        "\n",
        "\n",
        "Create a dataframe again (optional)...\n",
        "\n",
        "...once for our original data and then a copy for normalized data (just so you can see the difference here)\n",
        "\n",
        "\n",
        "\n",
        "```\n",
        "df_train = pd.DataFrame(x_train).add_prefix('c')\n",
        "df_test = pd.DataFrame(x_test).add_prefix('c')\n",
        "\n",
        "df_train_norm = df_train.copy()\n",
        "df_test_norm = df_test.copy()\n",
        "\n",
        "scaler = preprocessing.MinMaxScaler()\n",
        "data_scaler = scaler.fit(x_train[:,1:])\n",
        "\n",
        "\n",
        "```\n",
        "\n",
        "And apply it to our data, might seem a bit confusing, so let's see what I did here...\n",
        "\n",
        "We want to scale the data, not the classes, so we want columns 1 to 140 as 0 holds the classes. For this we use `.iloc[:,1:]` which does exactly what we want, it takes all the row indexes and column indexes 1 to 140. So that's what the left side is doing. \n",
        "\n",
        "Now to see what happens on the right side - we take our prepared scaler `data_scaler` use the `.transform()` on it to say we want to transform the data in the brackets. And as on the left side we say that we want to use data from the dataframe in columns 1 to 140 and by adding `.values` we get thir numerical values which we scale. \n",
        "\n",
        "Just like that our dataframe is updated with scaled values. \n",
        "\n",
        "\n",
        "```\n",
        "df_train_norm.iloc[:,1:] = data_scaler.transform(df_train_norm.iloc[:,1:].values)\n",
        "df_test_norm.iloc[:,1:] = data_scaler.transform(df_test_norm.iloc[:,1:].values)\n",
        "```\n",
        "\n",
        "Separate the data into normal and abnormal sets\n",
        "\n",
        "Well we know we have 5 classes, the first one with index 1 represents normal signals, whilst other 4 represents heart abnormalities.\n",
        "\n",
        "What we want is to have a class 0 that would represent the normal ECG signals and class 1 that holds all the abnormalities. Once again, pandas functions will help us with this. \n",
        "\n",
        "```\n",
        "\n",
        "\n",
        "df_train_normal = df_train_norm.query('c0 == 1')\n",
        "df_test_normal = df_train_norm.query('c0 == 1')\n",
        "\n",
        "df_train_abnormal = df_train_norm.query('c0 > 1')\n",
        "df_test_abnormal = df_train_norm.query('c0 > 1')\n",
        "```\n",
        "\n"
      ],
      "metadata": {
        "id": "RFzscfWyzmoD"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Let's look at our data and the difference"
      ],
      "metadata": {
        "id": "FWWC769Udwss"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "plt.plot(normal_x_train[0])\n",
        "plt.plot(normal_x_train[1])\n",
        "plt.plot(normal_x_train[2])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "K0ZUsQnKdhdP",
        "outputId": "eb6af1dc-8f66-485e-97da-7b976acc6e90"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f13091503d0>]"
            ]
          },
          "metadata": {},
          "execution_count": 14
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.plot(abnormal_x_train[0])\n",
        "plt.plot(abnormal_x_train[1])\n",
        "plt.plot(abnormal_x_train[2])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "hK11hvRkdnVr",
        "outputId": "ff1e4071-543b-43f5-da7c-15807c51908a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f130797b110>]"
            ]
          },
          "metadata": {},
          "execution_count": 15
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Create the autoencoder"
      ],
      "metadata": {
        "id": "FrfNwgPcd08z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class AutoEncoderDENSE(Model):\n",
        "    def __init__(self):\n",
        "        super(AutoEncoderDENSE, self).__init__()\n",
        "        self.encoder = tf.keras.Sequential([\n",
        "            tf.keras.layers.Dense(64, activation='relu'),\n",
        "            tf.keras.layers.Dense(32, activation='relu'),\n",
        "            tf.keras.layers.Dense(16, activation='relu'),\n",
        "            tf.keras.layers.Dense(8, activation='relu'),])\n",
        "        \n",
        "        self.decoder = tf.keras.Sequential([\n",
        "            tf.keras.layers.Dense(16, activation='relu'),\n",
        "            tf.keras.layers.Dense(32, activation='relu'),\n",
        "            tf.keras.layers.Dense(64, activation='relu'),\n",
        "            tf.keras.layers.Dense(140, activation='sigmoid'),])\n",
        "        \n",
        "    def call(self, x):\n",
        "        encoded = self.encoder(x)\n",
        "        decoded = self.decoder(encoded)\n",
        "        return decoded"
      ],
      "metadata": {
        "id": "qu4vz98SeN6A"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = AutoEncoderDENSE()\n",
        "early_stopping = tf.keras.callbacks.EarlyStopping(\n",
        "                    monitor='val_loss', \n",
        "                    patience=3,\n",
        "                    mode='min',)\n",
        "\n",
        "model.compile(optimizer='adam', loss='mae')"
      ],
      "metadata": {
        "id": "F4AZtvi2eqHd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "history = model.fit(normal_x_train, normal_x_train, \n",
        "                    epochs=50,\n",
        "                    batch_size=128,\n",
        "                    validation_data=(x_train[:,1:], x_train[:,1:]),\n",
        "                    callbacks=[early_stopping])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SPLsPtbjf2Kh",
        "outputId": "381de6bb-1c69-4a34-9e36-7cc8c224e9d7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "19/19 [==============================] - 3s 39ms/step - loss: 0.1144 - val_loss: 0.1017\n",
            "Epoch 2/50\n",
            "19/19 [==============================] - 0s 17ms/step - loss: 0.0699 - val_loss: 0.0803\n",
            "Epoch 3/50\n",
            "19/19 [==============================] - 0s 19ms/step - loss: 0.0505 - val_loss: 0.0763\n",
            "Epoch 4/50\n",
            "19/19 [==============================] - 0s 20ms/step - loss: 0.0469 - val_loss: 0.0755\n",
            "Epoch 5/50\n",
            "19/19 [==============================] - 0s 22ms/step - loss: 0.0465 - val_loss: 0.0751\n",
            "Epoch 6/50\n",
            "19/19 [==============================] - 0s 17ms/step - loss: 0.0464 - val_loss: 0.0752\n",
            "Epoch 7/50\n",
            "19/19 [==============================] - 0s 18ms/step - loss: 0.0463 - val_loss: 0.0751\n",
            "Epoch 8/50\n",
            "19/19 [==============================] - 0s 22ms/step - loss: 0.0462 - val_loss: 0.0743\n",
            "Epoch 9/50\n",
            "19/19 [==============================] - 0s 21ms/step - loss: 0.0461 - val_loss: 0.0743\n",
            "Epoch 10/50\n",
            "19/19 [==============================] - 0s 18ms/step - loss: 0.0460 - val_loss: 0.0737\n",
            "Epoch 11/50\n",
            "19/19 [==============================] - 0s 18ms/step - loss: 0.0460 - val_loss: 0.0732\n",
            "Epoch 12/50\n",
            "19/19 [==============================] - 0s 16ms/step - loss: 0.0459 - val_loss: 0.0730\n",
            "Epoch 13/50\n",
            "19/19 [==============================] - 0s 19ms/step - loss: 0.0458 - val_loss: 0.0728\n",
            "Epoch 14/50\n",
            "19/19 [==============================] - 0s 17ms/step - loss: 0.0456 - val_loss: 0.0728\n",
            "Epoch 15/50\n",
            "19/19 [==============================] - 0s 17ms/step - loss: 0.0455 - val_loss: 0.0723\n",
            "Epoch 16/50\n",
            "19/19 [==============================] - 0s 17ms/step - loss: 0.0450 - val_loss: 0.0716\n",
            "Epoch 17/50\n",
            "19/19 [==============================] - 0s 16ms/step - loss: 0.0438 - val_loss: 0.0704\n",
            "Epoch 18/50\n",
            "19/19 [==============================] - 0s 18ms/step - loss: 0.0409 - val_loss: 0.0670\n",
            "Epoch 19/50\n",
            "19/19 [==============================] - 0s 15ms/step - loss: 0.0379 - val_loss: 0.0623\n",
            "Epoch 20/50\n",
            "19/19 [==============================] - 0s 13ms/step - loss: 0.0364 - val_loss: 0.0609\n",
            "Epoch 21/50\n",
            "19/19 [==============================] - 0s 19ms/step - loss: 0.0358 - val_loss: 0.0595\n",
            "Epoch 22/50\n",
            "19/19 [==============================] - 0s 15ms/step - loss: 0.0354 - val_loss: 0.0593\n",
            "Epoch 23/50\n",
            "19/19 [==============================] - 0s 15ms/step - loss: 0.0349 - val_loss: 0.0585\n",
            "Epoch 24/50\n",
            "19/19 [==============================] - 0s 19ms/step - loss: 0.0346 - val_loss: 0.0582\n",
            "Epoch 25/50\n",
            "19/19 [==============================] - 0s 16ms/step - loss: 0.0342 - val_loss: 0.0585\n",
            "Epoch 26/50\n",
            "19/19 [==============================] - 0s 19ms/step - loss: 0.0341 - val_loss: 0.0577\n",
            "Epoch 27/50\n",
            "19/19 [==============================] - 0s 12ms/step - loss: 0.0336 - val_loss: 0.0575\n",
            "Epoch 28/50\n",
            "19/19 [==============================] - 0s 10ms/step - loss: 0.0334 - val_loss: 0.0573\n",
            "Epoch 29/50\n",
            "19/19 [==============================] - 0s 15ms/step - loss: 0.0332 - val_loss: 0.0578\n",
            "Epoch 30/50\n",
            "19/19 [==============================] - 0s 11ms/step - loss: 0.0332 - val_loss: 0.0567\n",
            "Epoch 31/50\n",
            "19/19 [==============================] - 0s 16ms/step - loss: 0.0328 - val_loss: 0.0564\n",
            "Epoch 32/50\n",
            "19/19 [==============================] - 0s 16ms/step - loss: 0.0327 - val_loss: 0.0567\n",
            "Epoch 33/50\n",
            "19/19 [==============================] - 0s 13ms/step - loss: 0.0325 - val_loss: 0.0568\n",
            "Epoch 34/50\n",
            "19/19 [==============================] - 0s 12ms/step - loss: 0.0322 - val_loss: 0.0565\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## So what did our model learn?\n",
        "\n",
        "The autoencoder is made of dense layers is trying to learn the distribution of the signal, so it can reproduce it at it's output. \n",
        "\n",
        "Since we are training it on the normal data it will learn to reproduce the normal data and will struggle with the abnormalities. "
      ],
      "metadata": {
        "id": "KOj9SUCpErM0"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We pass the normal test data into the model and plot their distribution in comparisson to their original shape. "
      ],
      "metadata": {
        "id": "ZRxIger8FUGf"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "decoder_out = model(normal_x_test).numpy()\n",
        "\n",
        "plt.plot(normal_x_test[1], 'r')\n",
        "plt.plot(decoder_out[1], 'b')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "MnVZsNaEChzh",
        "outputId": "e4ea4604-2b59-4fc6-8ff2-fe71a6f963b8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f12fe615690>]"
            ]
          },
          "metadata": {},
          "execution_count": 19
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Now we do the same thing, but with the abnormal data."
      ],
      "metadata": {
        "id": "BPMIGww0FfGb"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "decoder_out_abnormal = model(abnormal_x_test).numpy()\n",
        "\n",
        "plt.plot(abnormal_x_test[1], 'r')\n",
        "plt.plot(decoder_out_abnormal[1], 'b')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "2KWfnTuHEh1f",
        "outputId": "837a45b5-cfa4-487f-a6a8-2fcb06916912"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f130350a890>]"
            ]
          },
          "metadata": {},
          "execution_count": 20
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Finding threshold\n",
        "\n",
        "As we can see, the normal data is fitted pretty accurately, whilst the abnormal data does not really copy the original data. \n",
        "\n",
        "And that's what we are focusing at. We can now compute the mean square error `mse` or mean absolute error `mae` of the normal and abnormal data, plot their distribution and based on that decide where the threshold should be, so let's get into it!"
      ],
      "metadata": {
        "id": "fayvioyTgXLk"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Normal train data\n"
      ],
      "metadata": {
        "id": "thMr8TTxIGY8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "reconstruction = model.predict(normal_x_train)\n",
        "train_loss = tf.keras.losses.mse(reconstruction, normal_x_train) "
      ],
      "metadata": {
        "id": "CW3i7bPNgyK7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Plot the normal data distribution"
      ],
      "metadata": {
        "id": "RRXcLY65Gz3Y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "plt.hist(train_loss, bins=50)\n",
        "plt.title('Distribution of normal train data')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "AWffCdsXG4B2",
        "outputId": "3b579c5d-3fd2-4533-a038-e81f968f6723"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEICAYAAACktLTqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAUpElEQVR4nO3de5DlZX3n8fcnjEAQBYUJwnDpIbJGMGXpziIpY3RFNyJGyEYNtYmiyxZlxZgLuHFUskE3MZjsihoTLVZc8bIRZXWdBM2uEV2jq5gBIQjEOCCXGW7DyP1imPjdP84z5NB0T5/uc7qn++H9qurq3/X5fZ/zO/M5v/P8Tp9JVSFJ6suP7eoCJEmTZ7hLUocMd0nqkOEuSR0y3CWpQ4a7JHXIcO9Akg8m+d0JtXVoknuT7Nbmv5LkP0yi7dbeF5KcPKn25nHc309ye5JblvrY40jy2iRfW+RjXJnkBRNqa6LPFy2c4b7MJbkuyQNJ7klyZ5L/l+T1SR4+d1X1+qr6zyO29aKdbVNVN1TV3lX1TxOo/cwkH5/W/nFVdd64bc+zjkOB04Ejq+opS3nsxZRkKkklWTVOO1V1VFV9ZUJljWyU56MWznBfGX6hqp4AHAacBbwZOHfSBxk3JJaxQ4FtVXXbYh9ouT2Gy60eLR3DfQWpqruqagPwy8DJSZ4BkOQjSX6/Te+f5C/bVf4PkvxNkh9L8jEGIfcXbdjld4au/E5JcgNw0SxXgz+Z5FtJ7k7yuSRPbsd6QZLNwzXuuBpL8hLgrcAvt+Nd3tY//La91XVGkuuT3Jbko0n2aet21HFykhvakMrbZntskuzT9t/a2jujtf8i4IvAQa2Oj8yw7wuSbE5yeqvj5iSvm6vttu61Sb6e5Owk24Az2/n4szYEdW9b/5Qk70lyR5K/T/KsofbXJ7mmvTu7KskvjvaM4Kvt953tOD8zSz0/meSiJNva4/iJJPtOP2dt+swkn2r9vSeDIZt1O3ncX9z6c1eS9wMZWjfrcWd6Prbln05yS2vvq0mOGvGx0DSG+wpUVd8CNgPPm2H16W3dauAABgFbVfVq4AYG7wL2rqo/Gtrn+cDTgZ+f5ZCvAf49cCCwHXjfCDX+FfBO4Px2vGfOsNlr28+/Bg4H9gbeP22bnwWeBhwL/KckT5/lkH8C7NPaeX6r+XVV9dfAccBNrY7XzrL/U9r+a4BTgD9N8qSdtT2073OAaxk83n/Qlr0KOAPYH/gh8A3g0jZ/AfDuof2vYXAu9wHeDnw8yYGz1Dns59rvfVvfvjFLPQH+EDiIwXk+BDhzJ+2+HPgksC+wgUefE2BwIQF8Zqif1wDPHd5ktuPu5Pn4BeAI4CcYPF6f2PlDoNkY7ivXTcCTZ1j+EIMQPqyqHqqqv6m5v0DozKq6r6oemGX9x6rqO1V1H/C7wKvSbriO6VeAd1fVtVV1L/AW4KRp7xreXlUPVNXlwOXAo14kWi0nAW+pqnuq6jrgvwKvnkctDwHvaI/Z54F7gaeN2PZNVfUnVbV96DH8bFVdUlUPAp8FHqyqj7Z7GecDD1+5V9Wnq+qmqvpRVZ0PfA84eh61T/eIeqpqU1V9sap+WFVbGbywPH8n+3+tqj7fav0YMzzmzUuBK6vqgqp6CHgP8PAN6wUcl6r6cHucf8jgheCZO97NaX4M95VrDfCDGZb/MbAJ+D9Jrk2yfoS2bpzH+uuBxzG4UhvXQa294bZXMbji3GH40y33M7i6n27/VtP0ttbMo5ZtVbV9hmON0vZMj9+tQ9MPzDD/cD+SvCbJZRkMpd0JPIPxHt9H1JPkgCSfTLIlyd3Ax+dof/pjvmdmHrs/aPhY7SLi4fn5HjfJbknOakNUdwPXtVWTeK495hjuK1CSf8UgXB71Ebl21XN6VR3O4O31aUmO3bF6libnurI/ZGj6UAZXubcD9wF7DdW1G4PhoFHbvYnBTeLhtrfzyCAcxe2tpultbZlnOwtte8FfrZrkMOC/Ab8O7FdV+wLfYWjseidGPZ/vbMt+uqqeCPzqiO3P5WaGnhtJwiOfK3Mdd3qd/w44AXgRgyGqqR1NT6DWxxzDfQVJ8sQkL2MwHvrxqrpihm1eluSp7R/aXcA/AT9qq29lMG48X7+a5MgkewHvAC5ob9n/gcFV3fFJHsdg7HWPof1uBaYy9LHNaf4c+O0ka5PszT+P0W+fZfsZtVo+BfxBkie0wDyNwZXiWBaz7ebxDEJuK0C7kfuMEffdyuDcznVOn8BgmOmuJGuA/7iwUh/lQuCoJP+2Xdn/BoN7F6Med/rz8QkM7k9sY3DR8M4J1fmYZLivDH+R5B4Gb3nfxmDs8nWzbHsE8NcM/lF9A/izqvpyW/eHwBnt7f+b5nH8jwEfYfB2fU8G/4ipqruAXwM+xOBK9j4GN3N3+HT7vS3JpTO0++HW9leB7wMPAm+cR13D3tiOfy2DdzT/o7U/CYvWdlVdxWAM/xsMwu6nga+PuO/9DG6Yfr2d02Nm2fTtwLMZvNhfyOAm6Niq6nbglQw+nruNwXNvuPa5jjv9+fhRBkNeW4CrgG9Oos7HqvifdUhSf7xyl6QOGe6S1CHDXZI6ZLhLUoeWxZcK7b///jU1NbWry5CkFeWSSy65vapWz7RuWYT71NQUGzdu3NVlSNKKkuT62dY5LCNJHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR1aFn+hOo6p9RfOuPy6s45f4kokafnwyl2SOmS4S1KHDHdJ6pDhLkkdMtwlqUOGuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR0y3CWpQ4a7JHXIcJekDhnuktQhw12SOjRSuCf57SRXJvlOkj9PsmeStUkuTrIpyflJdm/b7tHmN7X1U4vZAUnSo80Z7knWAL8BrKuqZwC7AScB7wLOrqqnAncAp7RdTgHuaMvPbttJkpbQqMMyq4AfT7IK2Au4GXghcEFbfx5wYps+oc3T1h+bJJMpV5I0ijnDvaq2AP8FuIFBqN8FXALcWVXb22abgTVteg1wY9t3e9t+v+ntJjk1ycYkG7du3TpuPyRJQ0YZlnkSg6vxtcBBwOOBl4x74Ko6p6rWVdW61atXj9ucJGnIKMMyLwK+X1Vbq+oh4DPAc4F92zANwMHAlja9BTgEoK3fB9g20aolSTs1SrjfAByTZK82dn4scBXwZeAVbZuTgc+16Q1tnrb+oqqqyZUsSZrLKGPuFzO4MXopcEXb5xzgzcBpSTYxGFM/t+1yLrBfW34asH4R6pYk7cSquTeBqvo94PemLb4WOHqGbR8EXjl+aZKkhfIvVCWpQ4a7JHXIcJekDhnuktQhw12SOmS4S1KHDHdJ6pDhLkkdMtwlqUOGuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR0y3CWpQ4a7JHXIcJekDhnuktQhw12SOrRqVxewWKbWXzjj8uvOOn6JK5GkpeeVuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SerQSOGeZN8kFyT5+yRXJ/mZJE9O8sUk32u/n9S2TZL3JdmU5O+SPHtxuyBJmm7UK/f3An9VVT8FPBO4GlgPfKmqjgC+1OYBjgOOaD+nAh+YaMWSpDnNGe5J9gF+DjgXoKr+saruBE4AzmubnQec2KZPAD5aA98E9k1y4MQrlyTNapQr97XAVuC/J/l2kg8leTxwQFXd3La5BTigTa8Bbhzaf3Nb9ghJTk2yMcnGrVu3LrwHkqRHGSXcVwHPBj5QVc8C7uOfh2AAqKoCaj4HrqpzqmpdVa1bvXr1fHaVJM1hlHDfDGyuqovb/AUMwv7WHcMt7fdtbf0W4JCh/Q9uyyRJS2TOcK+qW4AbkzytLToWuArYAJzclp0MfK5NbwBe0z41cwxw19DwjSRpCYz6rZBvBD6RZHfgWuB1DF4YPpXkFOB64FVt288DLwU2Afe3bSVJS2ikcK+qy4B1M6w6doZtC3jDmHVJksbgX6hKUocMd0nqkOEuSR0y3CWpQ4a7JHXIcJekDhnuktQhw12SOmS4S1KHDHdJ6pDhLkkdMtwlqUOGuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR0y3CWpQ4a7JHXIcJekDhnuktQhw12SOmS4S1KHDHdJ6pDhLkkdGjnck+yW5NtJ/rLNr01ycZJNSc5Psntbvkeb39TWTy1O6ZKk2cznyv03gauH5t8FnF1VTwXuAE5py08B7mjLz27bSZKW0EjhnuRg4HjgQ20+wAuBC9om5wEntukT2jxt/bFte0nSEhn1yv09wO8AP2rz+wF3VtX2Nr8ZWNOm1wA3ArT1d7XtHyHJqUk2Jtm4devWBZYvSZrJnOGe5GXAbVV1ySQPXFXnVNW6qlq3evXqSTYtSY95q0bY5rnAy5O8FNgTeCLwXmDfJKva1fnBwJa2/RbgEGBzklXAPsC2iVcuSZrVnFfuVfWWqjq4qqaAk4CLqupXgC8Dr2ibnQx8rk1vaPO09RdVVU20aknSTo3zOfc3A6cl2cRgTP3ctvxcYL+2/DRg/XglSpLma5RhmYdV1VeAr7Tpa4GjZ9jmQeCVE6hNkrRA/oWqJHXIcJekDhnuktQhw12SOmS4S1KHDHdJ6pDhLkkdMtwlqUOGuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdWhe/xNTD6bWXzjruuvOOn4JK5GkxeOVuyR1yHCXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR0y3CWpQ4a7JHVoznBPckiSLye5KsmVSX6zLX9yki8m+V77/aS2PEnel2RTkr9L8uzF7oQk6ZFGuXLfDpxeVUcCxwBvSHIksB74UlUdAXypzQMcBxzRfk4FPjDxqiVJOzVnuFfVzVV1aZu+B7gaWAOcAJzXNjsPOLFNnwB8tAa+Ceyb5MCJVy5JmtW8xtyTTAHPAi4GDqiqm9uqW4AD2vQa4Mah3Ta3ZdPbOjXJxiQbt27dOs+yJUk7M3K4J9kb+J/Ab1XV3cPrqqqAms+Bq+qcqlpXVetWr149n10lSXMYKdyTPI5BsH+iqj7TFt+6Y7il/b6tLd8CHDK0+8FtmSRpiYzyaZkA5wJXV9W7h1ZtAE5u0ycDnxta/pr2qZljgLuGhm8kSUtg1QjbPBd4NXBFksvasrcCZwGfSnIKcD3wqrbu88BLgU3A/cDrJlqxJGlOc4Z7VX0NyCyrj51h+wLeMGZdkqQx+BeqktQhw12SOmS4S1KHDHdJ6pDhLkkdGuWjkI8ZU+svnHH5dWcdv8SVSNJ4vHKXpA4Z7pLUIcNdkjpkuEtShwx3SeqQ4S5JHTLcJalDhrskdchwl6QOGe6S1CHDXZI6ZLhLUocMd0nqkOEuSR3yK39H4FcBS1ppvHKXpA4Z7pLUIcNdkjpkuEtSh7yhOgZvtEparrxyl6QOGe6S1CHDXZI65Jj7InAsXtKu5pW7JHXIcJekDhnuktQhx9yXkGPxkpaKV+6S1CGv3JcBr+glTdqihHuSlwDvBXYDPlRVZy3GcXo3W+jPxhcDSTtMPNyT7Ab8KfBiYDPwt0k2VNVVkz6WHmm+7wB29uLhC4W0si3GlfvRwKaquhYgySeBEwDDfReZ7zuAhe4zCfN9IZrvi5BDYHqsWIxwXwPcODS/GXjO9I2SnAqc2mbvTfLdBRxrf+D2Bey3UnXf37zrEbNz9nfa9pM67q7S/fmd5rHWX5h8nw+bbcUuu6FaVecA54zTRpKNVbVuQiUte/a3b/a3f0vZ58X4KOQW4JCh+YPbMknSElmMcP9b4Igka5PsDpwEbFiE40iSZjHxYZmq2p7k14H/zeCjkB+uqisnfZxmrGGdFcj+9s3+9m/J+pyqWqpjSZKWiF8/IEkdMtwlqUPLMtyTvCTJd5NsSrJ+hvV7JDm/rb84ydTQure05d9N8vNLWfc4FtrnJPsl+XKSe5O8f6nrXqgx+vviJJckuaL9fuFS174QY/T36CSXtZ/Lk/ziUte+EOP8G27rD23P6TctVc3jGOP8TiV5YOgcf3BiRVXVsvphcBP2GuBwYHfgcuDIadv8GvDBNn0ScH6bPrJtvwewtrWz267u0yL3+fHAzwKvB96/q/uyBP19FnBQm34GsGVX92eR+7sXsKpNHwjctmN+uf6M09+h9RcAnwbetKv7s8jndwr4zmLUtRyv3B/++oKq+kdgx9cXDDsBOK9NXwAcmyRt+Ser6odV9X1gU2tvuVtwn6vqvqr6GvDg0pU7tnH6++2quqktvxL48SR7LEnVCzdOf++vqu1t+Z7ASvgExDj/hklyIvB9Bud3JRirv4tlOYb7TF9fsGa2bdoT/y5gvxH3XY7G6fNKNKn+/hJwaVX9cJHqnJSx+pvkOUmuBK4AXj8U9svVgvubZG/gzcDbl6DOSRn3+bw2ybeT/N8kz5tUUX6fu1akJEcB7wL+za6uZbFV1cXAUUmeDpyX5AtVtZLeqc3HmcDZVXXvIl/YLhc3A4dW1bYk/xL4X0mOqqq7x214OV65j/L1BQ9vk2QVsA+wbcR9l6Nx+rwSjdXfJAcDnwVeU1XXLHq145vI+a2qq4F7GdxrWM7G6e9zgD9Kch3wW8Bb2x9FLmcL7m8bQt4GUFWXMBi7/xeTKGo5hvsoX1+wATi5Tb8CuKgGdyc2ACe1O9NrgSOAby1R3eMYp88r0YL7m2Rf4EJgfVV9fckqHs84/V3bwoAkhwE/BVy3NGUv2IL7W1XPq6qpqpoC3gO8s6qW+6fAxjm/qzP4PzBIcjiDzLp2IlXt6jvNs9x9finwDwxexd7Wlr0DeHmb3pPBnfRNDML78KF939b2+y5w3K7uyxL1+TrgBwyu6jYz7U79cvxZaH+BM4D7gMuGfn5iV/dnEfv7agY3Fi8DLgVO3NV9Wcz+TmvjTFbAp2XGPL+/NO38/sKkavLrBySpQ8txWEaSNCbDXZI6ZLhLUocMd0nqkOEuSR0y3CWpQ4a7JHXo/wMGCGuPqHzSjwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Get the mean value and standard deviation of the data"
      ],
      "metadata": {
        "id": "TOXyof8AHU1L"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "mean_normal = np.mean(train_loss)\n",
        "mean_normal"
      ],
      "metadata": {
        "id": "LmDJQIBChLzo",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0414a552-3c6f-43c2-e4df-589b123650d8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.0024309782558374488"
            ]
          },
          "metadata": {},
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "std_normal = np.std(train_loss)\n",
        "std_normal"
      ],
      "metadata": {
        "id": "vLCs050shR0I",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0dbd83ba-cfb0-49fb-f7b6-93a038d76152"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.0030271994922538795"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let's decide where our threshold is going to be...\n",
        "\n",
        "We can take the mean value + 2*std and see what we get. "
      ],
      "metadata": {
        "id": "kN8KMa85HkEa"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "threshold = mean_normal + 2*std_normal\n",
        "threshold"
      ],
      "metadata": {
        "id": "lE4POP78ha4k",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9868b61b-7f26-4023-d506-d19681fb9c94"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.008485377240345208"
            ]
          },
          "metadata": {},
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "If we look at our histogram, 0.064 seem like a good threshold, where most of our normal data lies on the left side. Now we should see, how the distribution of the abnormal data looks like."
      ],
      "metadata": {
        "id": "lVSR5IfnH1LD"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Evaluation"
      ],
      "metadata": {
        "id": "20ufFd_fT7-5"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Normal data"
      ],
      "metadata": {
        "id": "N6PEzvUXT-51"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Calculate the loss on the testing data and plot the histogram"
      ],
      "metadata": {
        "id": "aDzDqcnoUaUR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "reconstruction_test = model.predict(normal_x_test)\n",
        "test_loss_normal = tf.keras.losses.mse(reconstruction_test, normal_x_test) "
      ],
      "metadata": {
        "id": "CwaYxGAQT-BK"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plt.hist(test_loss_normal, bins=50)\n",
        "plt.title('Distribution of normal test data')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "t2hTXgWKUZLF",
        "outputId": "851a0b5d-6193-4e2d-b742-d2e47232a1eb"
      },
      "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": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Abnormal data"
      ],
      "metadata": {
        "id": "ceALq6aYIIyL"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Again, compute the MAE of the predicted abnormal data and display it in the histogram."
      ],
      "metadata": {
        "id": "4HlIzz2qIQxI"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "reconstruction_anomalies = model.predict(abnormal_x_test)\n",
        "test_loss_anomalies = tf.keras.losses.mse(reconstruction_anomalies, abnormal_x_test)"
      ],
      "metadata": {
        "id": "MA8CEqlxhf1E"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plt.hist(test_loss_anomalies, bins=50)\n",
        "plt.title('Distribution of abnormal test data')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "UntiWMOYIOIQ",
        "outputId": "03b27573-5143-4fb8-a2ea-3a34e2c6d149"
      },
      "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": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Seems like our threshold should work well in detecting the most abnormalities. So let's put it all together."
      ],
      "metadata": {
        "id": "67lGzpsnIYZ0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# plot histograms of normal and abnormal data\n",
        "plt.hist(test_loss_normal, bins=50, label='normal')\n",
        "plt.hist(test_loss_anomalies, bins=50, label='anomalies')\n",
        "# plot a vertical line, which displays the threshold\n",
        "plt.axvline(threshold, color='r', linewidth=3, linestyle='dashed', label='{:0.3f}'.format(threshold))\n",
        "# add a legend and title\n",
        "plt.legend(loc='upper right')\n",
        "plt.title('Threshold for detecting anomalies')\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "oYdommTvhzcW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "outputId": "d54e07ae-2249-403f-ff95-e14dbf2b3c51"
      },
      "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": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Confusion matrix, ROC curve"
      ],
      "metadata": {
        "id": "JlTD3i5kSz55"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "To be able to plot the confusion matrix and ROC curve we first need to count how many errors I. and II. we have -> how many false positives or false negatives. \n",
        "\n",
        "For this we are going to use the tf.math function, where we compare the threshold with the test losses of normal and abnormal data and get an array with True/False values. \n",
        "\n",
        "Next we count how many manu nonzero values we have (non-zero = 1 = True)."
      ],
      "metadata": {
        "id": "_UPXiNGQVDoJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# We correctly detect normal data if the normal loss is smaller than the threshold\n",
        "pred_normal = tf.math.less(test_loss_normal, threshold)\n",
        "# We correctly detect abnormal data if the normal loss is greater than the threshold\n",
        "pred_anomaly = tf.math.greater(test_loss_anomalies, threshold)\n"
      ],
      "metadata": {
        "id": "fDoD58YO5jxm"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "pred_normal_true = tf.math.count_nonzero(pred_normal).numpy()\n",
        "pred_normal_false = normal_x_test.shape[0] - pred_normal_true\n",
        "pred_normal_true, pred_normal_false"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1T4BoheU5pMO",
        "outputId": "cf44e191-2694-4d80-e9c9-b6c968008133"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(562, 24)"
            ]
          },
          "metadata": {},
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pred_anomaly_true = tf.math.count_nonzero(pred_anomaly).numpy()\n",
        "pred_anomaly_false = abnormal_x_test.shape[0] - pred_anomaly_true\n",
        "pred_anomaly_true, pred_anomaly_false"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "g28zUtz554rH",
        "outputId": "d0deb039-7f7f-4e64-9574-bf6ad6257423"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(408, 6)"
            ]
          },
          "metadata": {},
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "To **plot the confusion matrix**, we need to create a list which will hold the values we want to display. \n",
        "\n",
        "We can also define the categories to display as ticks - normal/anomalities. \n",
        "\n",
        "And to plot the data we will use **seaborn**. Seaborn is a library for statistical data visualization and is based on matplotlib, but is more user friendly.\n",
        "Basicly all we could do is to write sns.heatmap(cm) and we would have our confusion matrix. However we added few extras to make it nice and representable. "
      ],
      "metadata": {
        "id": "-57kJkMjgECx"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cm = [[pred_normal_true, pred_normal_false], \n",
        "      [pred_anomaly_false, pred_anomaly_true]]\n",
        "\n",
        "# cm = [[pred_anomaly_true, pred_anomaly_false]\n",
        "#       [pred_normal_false, pred_normal_true]]\n",
        "\n",
        "categories = ['Normal', 'Anomalies']\n",
        "\n",
        "g = sns.heatmap(cm, annot=True, fmt=\"d\",\n",
        "            xticklabels=categories, \n",
        "            yticklabels=categories, \n",
        "            cbar=False, \n",
        "            cmap=\"GnBu\")\n",
        "\n",
        "g.set(xlabel='Predicted', ylabel='Actual', title='Confusion Matrix')\n",
        "\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "lgIdL1Mk5w1n",
        "outputId": "050d45e8-a738-40ac-82cc-b17624d56fa6"
      },
      "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": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Now we combine the first three cells of this block into one as a function which will predict true positives, false negatives, true negatives and false positives. We will need this function to get us values, so we can plot ROC."
      ],
      "metadata": {
        "id": "EncIAGvavseV"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def predictor(normal, abnormal, threshold):\n",
        "   \n",
        "    # We correctly detect normal data if the normal loss is smaller than the threshold\n",
        "    pred_normal = tf.math.less(normal, threshold)\n",
        "    # We correctly detect abnormal data if the normal loss is greater than the threshold\n",
        "    pred_anomaly = tf.math.greater(abnormal, threshold)\n",
        "\n",
        "    tn = tf.math.count_nonzero(pred_normal).numpy()\n",
        "    fp = normal.shape[0] - tn\n",
        "\n",
        "    tp = tf.math.count_nonzero(pred_anomaly).numpy()\n",
        "    fn = abnormal.shape[0] - tp\n",
        "    \n",
        "    return tp,tn,fp,fn"
      ],
      "metadata": {
        "id": "mLePzSefjk_H"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Calculate the roc values - tpr and fpr...\n",
        "\n",
        "...first define empty list which will hold the values\n",
        "\n",
        "...then create a for loop which will take 100 values from 0 to 1 (use `np.linspace(from, to, steps)`)\n",
        "\n",
        "...now use the defined predictor function and, where the threshold is the value 0 to 1 of the forloop\n",
        "\n",
        "...calculate the trp and frp for each iteration `tpr = tp/(tp + fn)` and `fpr = fp/(fp + tn)`\n",
        "\n",
        "...`append` the tpr and fpr values in the roc list"
      ],
      "metadata": {
        "id": "i_oFJtI5GoI-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "tpr_values = []\n",
        "fpr_values = []\n",
        "\n",
        "for thresh in np.linspace(0, 1, 100):\n",
        "    tp,tn,fp,fn = predictor(test_loss_normal, \n",
        "                            test_loss_anomalies, \n",
        "                            thresh/10)\n",
        "    tpr = tp/(tp + fn)\n",
        "    fpr = fp/(fp + tn)\n",
        "    tpr_values.append(tpr)\n",
        "    fpr_values.append(fpr)"
      ],
      "metadata": {
        "id": "5-vvgy4F58W_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Plot the ROC curve"
      ],
      "metadata": {
        "id": "CUGlz5TuyRU_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "fig, ax = plt.subplots(figsize=(10,7))\n",
        "ax.plot(fpr_values, tpr_values)\n",
        "ax.plot(np.linspace(0, 1, 100),\n",
        "         np.linspace(0, 1, 100),\n",
        "         label='baseline',\n",
        "         linestyle='--')\n",
        "plt.title('Receiver Operating Characteristic Curve', fontsize=18)\n",
        "plt.ylabel('TPR', fontsize=16)\n",
        "plt.xlabel('FPR', fontsize=16)\n",
        "plt.legend(fontsize=12);"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 467
        },
        "id": "hJfJ6ThdhigQ",
        "outputId": "4ec691ea-068b-491d-ff8c-a91a0ec4bc3d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}