binary_classification_svm.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "HKLF7zkiBkWo"
      },
      "outputs": [],
      "source": [
        "# Imports\n",
        "import numpy as np\n",
        "import pandas as pd \n",
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wBCFK-vpCJU1"
      },
      "source": [
        "# Loading Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "omJ02RGxCLqx",
        "outputId": "c8a50c3b-515b-4fde-dbe3-7f71cf3bf3cd"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n",
            "Train data shape: ((600, 4610), (2800, 4610))\n",
            "Test data shape: (2000, 4608)\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "train_data_1 = pd.read_csv('/content/drive/MyDrive/MachineLearning/Report/training1.csv')\n",
        "train_data_2 = pd.read_csv('/content/drive/MyDrive/MachineLearning/Report/training2.csv')\n",
        "test_data = pd.read_csv('/content/drive/MyDrive/MachineLearning/Report/test.csv')\n",
        "\n",
        "print(f'Train data shape: {train_data_1.shape, train_data_2.shape}')\n",
        "print(f'Test data shape: {test_data.shape}')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "U-pNT4aDPlLm",
        "outputId": "22e0ce2f-5d2e-4622-9a3e-30d146a5109c"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3400, 4610)"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ],
      "source": [
        "# Combine training datasets\n",
        "train_data = pd.concat([train_data_1, train_data_2])\n",
        "train_data.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "5qd29CeHVln5",
        "outputId": "c8706d25-0177-4762-8283-cba7b1b87c18"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "      CNNs   CNNs.1    CNNs.2    CNNs.3  CNNs.4   CNNs.5   CNNs.6   CNNs.7  \\\n",
              "0  0.96629  0.00000  0.000000  0.184890  1.3535  0.00000  0.66736  0.00000   \n",
              "1  0.42980  0.00000  0.142340  0.000000  0.0000  0.00000  0.00000  0.83749   \n",
              "2  0.00000  0.00000  0.000000  0.968570  0.0000  0.00000  0.00000  0.00000   \n",
              "3  0.00000  0.13278  0.032179  0.742990  0.0000  0.67521  0.00000  0.18638   \n",
              "4  0.00000  0.12242  0.627020  0.017635  0.0000  0.00000  0.28884  1.10050   \n",
              "\n",
              "    CNNs.8   CNNs.9  ...  GIST.504  GIST.505  GIST.506  GIST.507  GIST.508  \\\n",
              "0  0.30246  0.00000  ...  0.025722  0.029769  0.022448  0.018352  0.059976   \n",
              "1  0.00000  0.00000  ...  0.053810  0.030060  0.036972  0.039459  0.025726   \n",
              "2  0.00000  0.00000  ...  0.042560  0.045289  0.047827  0.039196  0.046091   \n",
              "3  1.06300  0.74294  ...  0.040872  0.035247  0.033755  0.049919  0.058322   \n",
              "4  0.00000  0.15511  ...  0.025219  0.005606  0.024671  0.025305  0.025466   \n",
              "\n",
              "   GIST.509  GIST.510  GIST.511  label  confidence  \n",
              "0  0.018121  0.034956  0.018626    0.0        1.00  \n",
              "1  0.015295  0.032943  0.015002    1.0        0.66  \n",
              "2  0.020280  0.048154  0.030391    1.0        1.00  \n",
              "3  0.027145  0.018040  0.030181    1.0        1.00  \n",
              "4  0.023212  0.011266  0.035695    0.0        0.66  \n",
              "\n",
              "[5 rows x 4610 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-d27d6d73-34af-4995-bbef-04cf51241466\">\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>CNNs</th>\n",
              "      <th>CNNs.1</th>\n",
              "      <th>CNNs.2</th>\n",
              "      <th>CNNs.3</th>\n",
              "      <th>CNNs.4</th>\n",
              "      <th>CNNs.5</th>\n",
              "      <th>CNNs.6</th>\n",
              "      <th>CNNs.7</th>\n",
              "      <th>CNNs.8</th>\n",
              "      <th>CNNs.9</th>\n",
              "      <th>...</th>\n",
              "      <th>GIST.504</th>\n",
              "      <th>GIST.505</th>\n",
              "      <th>GIST.506</th>\n",
              "      <th>GIST.507</th>\n",
              "      <th>GIST.508</th>\n",
              "      <th>GIST.509</th>\n",
              "      <th>GIST.510</th>\n",
              "      <th>GIST.511</th>\n",
              "      <th>label</th>\n",
              "      <th>confidence</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0.96629</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.184890</td>\n",
              "      <td>1.3535</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.66736</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.30246</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.025722</td>\n",
              "      <td>0.029769</td>\n",
              "      <td>0.022448</td>\n",
              "      <td>0.018352</td>\n",
              "      <td>0.059976</td>\n",
              "      <td>0.018121</td>\n",
              "      <td>0.034956</td>\n",
              "      <td>0.018626</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.42980</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.142340</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.83749</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.053810</td>\n",
              "      <td>0.030060</td>\n",
              "      <td>0.036972</td>\n",
              "      <td>0.039459</td>\n",
              "      <td>0.025726</td>\n",
              "      <td>0.015295</td>\n",
              "      <td>0.032943</td>\n",
              "      <td>0.015002</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.66</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.968570</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.042560</td>\n",
              "      <td>0.045289</td>\n",
              "      <td>0.047827</td>\n",
              "      <td>0.039196</td>\n",
              "      <td>0.046091</td>\n",
              "      <td>0.020280</td>\n",
              "      <td>0.048154</td>\n",
              "      <td>0.030391</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.13278</td>\n",
              "      <td>0.032179</td>\n",
              "      <td>0.742990</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.67521</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.18638</td>\n",
              "      <td>1.06300</td>\n",
              "      <td>0.74294</td>\n",
              "      <td>...</td>\n",
              "      <td>0.040872</td>\n",
              "      <td>0.035247</td>\n",
              "      <td>0.033755</td>\n",
              "      <td>0.049919</td>\n",
              "      <td>0.058322</td>\n",
              "      <td>0.027145</td>\n",
              "      <td>0.018040</td>\n",
              "      <td>0.030181</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.12242</td>\n",
              "      <td>0.627020</td>\n",
              "      <td>0.017635</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.28884</td>\n",
              "      <td>1.10050</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.15511</td>\n",
              "      <td>...</td>\n",
              "      <td>0.025219</td>\n",
              "      <td>0.005606</td>\n",
              "      <td>0.024671</td>\n",
              "      <td>0.025305</td>\n",
              "      <td>0.025466</td>\n",
              "      <td>0.023212</td>\n",
              "      <td>0.011266</td>\n",
              "      <td>0.035695</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.66</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 4610 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-d27d6d73-34af-4995-bbef-04cf51241466')\"\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-d27d6d73-34af-4995-bbef-04cf51241466 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-d27d6d73-34af-4995-bbef-04cf51241466');\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",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 4
        }
      ],
      "source": [
        "train_data.head()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "test_data.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "X_JwpEIGEgiJ",
        "outputId": "ce99308b-f404-4720-d57e-3ff1f8c7b252"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "       CNNs   CNNs.1  CNNs.2   CNNs.3   CNNs.4   CNNs.5   CNNs.6   CNNs.7  \\\n",
              "0  0.787990      NaN  2.0246  0.00000  0.00000  3.29550  0.00000  0.12965   \n",
              "1  0.000000  0.72265     NaN  0.15856      NaN      NaN  0.00000  0.00000   \n",
              "2  0.000000      NaN  0.0000  0.00000  0.64827  0.48099  0.00000  0.71721   \n",
              "3  0.044644  0.00000     NaN  0.71294  0.00000      NaN      NaN  0.00000   \n",
              "4  0.848170      NaN  0.0000      NaN  0.00000  0.00000  0.25626  0.00000   \n",
              "\n",
              "    CNNs.8  CNNs.9  ...  GIST.502  GIST.503  GIST.504  GIST.505  GIST.506  \\\n",
              "0  0.00000     NaN  ...       NaN  0.034861  0.045078       NaN       NaN   \n",
              "1      NaN     0.0  ...  0.004348  0.007177  0.014610       NaN  0.007715   \n",
              "2  0.70541     0.0  ...  0.026724  0.027204  0.021123  0.027027  0.029328   \n",
              "3  0.00000     0.0  ...  0.026242  0.029739  0.025212       NaN  0.021912   \n",
              "4  0.79886     0.0  ...  0.003493  0.002661       NaN  0.069194  0.040851   \n",
              "\n",
              "   GIST.507  GIST.508  GIST.509  GIST.510  GIST.511  \n",
              "0  0.031250  0.026715  0.007441  0.028426       NaN  \n",
              "1  0.014736  0.009790  0.036324       NaN       NaN  \n",
              "2  0.031330       NaN  0.013995  0.033780  0.013178  \n",
              "3  0.020740  0.024278  0.001928  0.015194  0.023487  \n",
              "4  0.006043       NaN  0.035340  0.024834  0.076113  \n",
              "\n",
              "[5 rows x 4608 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-66019415-e23a-445a-8d25-9cdadea97a5e\">\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>CNNs</th>\n",
              "      <th>CNNs.1</th>\n",
              "      <th>CNNs.2</th>\n",
              "      <th>CNNs.3</th>\n",
              "      <th>CNNs.4</th>\n",
              "      <th>CNNs.5</th>\n",
              "      <th>CNNs.6</th>\n",
              "      <th>CNNs.7</th>\n",
              "      <th>CNNs.8</th>\n",
              "      <th>CNNs.9</th>\n",
              "      <th>...</th>\n",
              "      <th>GIST.502</th>\n",
              "      <th>GIST.503</th>\n",
              "      <th>GIST.504</th>\n",
              "      <th>GIST.505</th>\n",
              "      <th>GIST.506</th>\n",
              "      <th>GIST.507</th>\n",
              "      <th>GIST.508</th>\n",
              "      <th>GIST.509</th>\n",
              "      <th>GIST.510</th>\n",
              "      <th>GIST.511</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0.787990</td>\n",
              "      <td>NaN</td>\n",
              "      <td>2.0246</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>3.29550</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.12965</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.034861</td>\n",
              "      <td>0.045078</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.031250</td>\n",
              "      <td>0.026715</td>\n",
              "      <td>0.007441</td>\n",
              "      <td>0.028426</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.72265</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.15856</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.004348</td>\n",
              "      <td>0.007177</td>\n",
              "      <td>0.014610</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.007715</td>\n",
              "      <td>0.014736</td>\n",
              "      <td>0.009790</td>\n",
              "      <td>0.036324</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.64827</td>\n",
              "      <td>0.48099</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.71721</td>\n",
              "      <td>0.70541</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.026724</td>\n",
              "      <td>0.027204</td>\n",
              "      <td>0.021123</td>\n",
              "      <td>0.027027</td>\n",
              "      <td>0.029328</td>\n",
              "      <td>0.031330</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.013995</td>\n",
              "      <td>0.033780</td>\n",
              "      <td>0.013178</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.044644</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.71294</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.026242</td>\n",
              "      <td>0.029739</td>\n",
              "      <td>0.025212</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.021912</td>\n",
              "      <td>0.020740</td>\n",
              "      <td>0.024278</td>\n",
              "      <td>0.001928</td>\n",
              "      <td>0.015194</td>\n",
              "      <td>0.023487</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.848170</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.25626</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.79886</td>\n",
              "      <td>0.0</td>\n",
              "      <td>...</td>\n",
              "      <td>0.003493</td>\n",
              "      <td>0.002661</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.069194</td>\n",
              "      <td>0.040851</td>\n",
              "      <td>0.006043</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.035340</td>\n",
              "      <td>0.024834</td>\n",
              "      <td>0.076113</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 4608 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-66019415-e23a-445a-8d25-9cdadea97a5e')\"\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-66019415-e23a-445a-8d25-9cdadea97a5e 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-66019415-e23a-445a-8d25-9cdadea97a5e');\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",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.rcParams['figure.figsize'] = [5, 5]\n",
        "\n",
        "# Check class balance\n",
        "balance = train_data['label'].value_counts()/len(train_data)\n",
        "plt.xticks(np.arange(0,2))\n",
        "plt.bar(x=[0,1], height=balance.values, color=['blue', 'orange'])\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 320
        },
        "id": "2IE6gYJ_OpLI",
        "outputId": "866bf240-ccd2-4506-8b11-1ed8eed50088"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nc58JMp456Hu"
      },
      "source": [
        "# Pre-processing"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6ap1LJhOPtfm"
      },
      "source": [
        "## Imputation"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hV228j8lUHtS",
        "outputId": "3e1440a6-e075-4965-ee92-c3d9ac67f51d"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "CNNs          545\n",
              "CNNs.1        583\n",
              "CNNs.2        584\n",
              "CNNs.3        528\n",
              "CNNs.4        556\n",
              "             ... \n",
              "GIST.509      540\n",
              "GIST.510      556\n",
              "GIST.511      568\n",
              "label           0\n",
              "confidence      0\n",
              "Length: 4610, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ],
      "source": [
        "# Print the number of missing values\n",
        "train_data.isna().sum()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1-OSdFZgWt1j",
        "outputId": "94a539bb-be36-4c3c-ab36-ecd361885c77"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3400, 4610)"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ],
      "source": [
        "# Remove feature if it has more than 20% of missing values\n",
        "train_data = train_data.dropna(thresh=train_data.shape[0]*0.80, axis=1)\n",
        "train_data.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qCdLlxvJgkRC",
        "outputId": "455f85e8-a9d4-4a31-fec3-1f16e91624bc"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ],
      "source": [
        "# Impute missing values with column means\n",
        "train_data = train_data.fillna(value=train_data.mean())\n",
        "\n",
        "# Check if any column has missing values\n",
        "num_nan = train_data.isna().sum().values\n",
        "len(num_nan[num_nan != 0])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7fXz0J5PxvSz"
      },
      "source": [
        "## Standartization"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 606
        },
        "id": "34mNrFiOK_4w",
        "outputId": "6df78095-52ad-450c-8e1e-f0156339745b"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x720 with 6 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "# Visualise distributions of 6 random features\n",
        "cols_1 = train_data.iloc[:,:-514].sample(n=3,axis='columns', random_state=0) # Take 3 CNN features\n",
        "cols_2 = train_data.iloc[:,-514:].sample(n=3,axis='columns', random_state=0) # And 3 gist features\n",
        "cols = pd.concat([cols_1, cols_2], axis=1)\n",
        "\n",
        "plt.rcParams['figure.figsize'] = [15, 10]\n",
        "for i, col in enumerate(cols):\n",
        "    plt.subplot(2,3,i+1)\n",
        "    ax = cols[col].hist(bins=16)\n",
        "    ax.set_xlabel(col)\n",
        "\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kU7UbuD9IxO5",
        "outputId": "a1290602-75f8-4fe3-e4cd-f2499559b945"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3400, 4610)"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ],
      "source": [
        "from sklearn.preprocessing import StandardScaler\n",
        "\n",
        "# Scale values to have a mean of 0 and std of 1\n",
        "std_scaler = StandardScaler()\n",
        "train_data_norm = std_scaler.fit_transform(train_data)\n",
        "train_data_norm = pd.DataFrame(train_data_norm, index=train_data.index, columns=train_data.columns) # convert back to DataFrame\n",
        "train_data_norm.shape"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Plot the distributions again to see what has changed\n",
        "cols_1 = train_data_norm.iloc[:,:-514].sample(n=3,axis='columns', random_state=0) # Take 3 CNN features\n",
        "cols_2 = train_data_norm.iloc[:,-514:-2].sample(n=3,axis='columns', random_state=0) # And 3 gist features\n",
        "cols = pd.concat([cols_1, cols_2], axis=1)\n",
        "\n",
        "plt.rcParams['figure.figsize'] = [15, 10]\n",
        "for i, col in enumerate(cols):\n",
        "    plt.subplot(2,3,i+1)\n",
        "    ax = cols[col].hist(bins=16)\n",
        "    ax.set_xlabel(col)\n",
        "\n",
        "plt.savefig(\"hists.png\", bbox_inches = 'tight')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 606
        },
        "id": "mnhIU5vH1FoD",
        "outputId": "8a46e07b-7a7c-41a3-bd84-34138b2d1ab0"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x720 with 6 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A8qNKDKh4qZJ"
      },
      "source": [
        "## Normalization"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xa61oPzn4sD_",
        "outputId": "7572677c-5304-4618-8733-40821835f08f"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3400, 4610)"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ],
      "source": [
        "from sklearn.preprocessing import MinMaxScaler \n",
        "\n",
        "# Translate features to range 0-1\n",
        "minmax_scaler = MinMaxScaler()\n",
        "train_data_norm = minmax_scaler.fit_transform(train_data_norm)\n",
        "train_data_norm = pd.DataFrame(train_data_norm, index=train_data.index, columns=train_data.columns) # convert back to DataFrame\n",
        "train_data_norm.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 606
        },
        "id": "zkGAHHdT5T25",
        "outputId": "7051683e-7dd5-42f5-990c-83238b3a1237"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1080x720 with 6 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "# Plot the distributions again to see what has changed\n",
        "cols_1 = train_data_norm.iloc[:,:-514].sample(n=3,axis='columns', random_state=0) # Take 3 CNN features\n",
        "cols_2 = train_data_norm.iloc[:,-514:-2].sample(n=3,axis='columns', random_state=0) # And 3 gist features\n",
        "cols = pd.concat([cols_1, cols_2], axis=1)\n",
        "\n",
        "plt.rcParams['figure.figsize'] = [15, 10]\n",
        "for i, col in enumerate(cols):\n",
        "    plt.subplot(2,3,i+1)\n",
        "    ax = cols[col].hist(bins=16)\n",
        "    ax.set_xlabel(col)\n",
        "\n",
        "plt.savefig(\"hists.png\", bbox_inches = 'tight')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BZwvywfB8jPP"
      },
      "source": [
        "# Model Selection"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Separate features from class labels\n",
        "data = train_data.drop(['label','confidence'], axis=1)\n",
        "classes = train_data['label']\n",
        "confidences = train_data['confidence']"
      ],
      "metadata": {
        "id": "TEy969d4zxN2"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "AG8w6mlfebE3"
      },
      "outputs": [],
      "source": [
        "from sklearn.svm import SVC\n",
        "from sklearn.pipeline import Pipeline"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "rdjuvFHLT2o3"
      },
      "outputs": [],
      "source": [
        "# Define classifiers\n",
        "classifiers = { \n",
        "    'svm_rbf': Pipeline([('std_scaler', std_scaler), \n",
        "                         ('minmax_scaler', minmax_scaler), \n",
        "                         ('clf', SVC(C=1.0, kernel='rbf'))]),\n",
        "               \n",
        "    'svm_linear': Pipeline([('std_scaler', std_scaler), \n",
        "                            ('minmax_scaler', minmax_scaler), \n",
        "                            ('clf', SVC(C=0.01, kernel='linear'))]),           \n",
        "    }"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OJO5fII_s9Wg"
      },
      "source": [
        "## Choosing Dimensionality Reduction Method"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MVejgTuitFd9"
      },
      "source": [
        "Use Anova to select most important features:"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "kPAQDmowaoK_"
      },
      "outputs": [],
      "source": [
        "from sklearn.feature_selection import SelectKBest, f_classif\n",
        "from sklearn.model_selection import cross_validate\n",
        "\n",
        "def do_feature_selection(clf, X, y, num_feats, label, include_plot=True):\n",
        "    '''\n",
        "    Perform Anova feature selection\n",
        "    :param clf: pipeline object with an estimator at the end\n",
        "    :X: dataset of observations\n",
        "    :y: corresponding class labels\n",
        "    :num_feats: list of numbers corresponding to the amount of features to be tried\n",
        "    :include_plot: True if the figure should be drawn after selection\n",
        "    :return: a dict where each key is the number of features and corresponding value is the list of selected features\n",
        "    '''\n",
        "    results = {}\n",
        "    acc = []\n",
        "\n",
        "    # Add anova as a first step of a pipeline\n",
        "    clf.steps.insert(0,['anova', SelectKBest(f_classif)])\n",
        "\n",
        "    for i in num_feats:\n",
        "        # Set the number of features for anova\n",
        "        clf.set_params(anova__k=i)\n",
        "\n",
        "        # Cross-validate\n",
        "        cv = cross_validate(clf, X, y, cv=5, return_estimator=True)\n",
        "        scores = cv['test_score'] # get the scores for validation data\n",
        "        estimators = cv['estimator'] # get the classifiers used during cross validation\n",
        "\n",
        "        # Save selected feature indexes\n",
        "        feats = estimators[0].steps[0][1].get_support(indices=True) \n",
        "        results[i] = feats \n",
        "        acc.append(np.mean(scores))\n",
        "\n",
        "    # Remove anova from the pipeline\n",
        "    clf.steps.pop(0)\n",
        "\n",
        "    # Plot the accuracy\n",
        "    if include_plot:\n",
        "        plt.plot(num_features, acc, label=label)\n",
        "        plt.legend(loc='lower right')\n",
        "    \n",
        "    return results"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Number of features for anova to be tried\n",
        "num_features = [144, 288, 576, 1152, 2304, 4608]\n",
        "\n",
        "plt.rcParams['figure.figsize'] = [5, 5]\n",
        "plt.title('ANOVA')\n",
        "plt.xlabel('Number of features selected')\n",
        "plt.ylabel('Accuracy')\n",
        "\n",
        "clf_results_anova = {}\n",
        "for label, clf in classifiers.items():\n",
        "    feats = do_feature_selection(clf, data, classes, num_features, label)\n",
        "    clf_results_anova[label] = feats\n",
        "\n",
        "plt.savefig(\"anova.png\", bbox_inches = 'tight')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 350
        },
        "id": "fplM60JtIrTq",
        "outputId": "ee82f8e2-f547-45e8-f8ba-7eff43c7b3ab"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iZEN0E4R4ncj"
      },
      "source": [
        "Use PCA to project features on a smaller number of dimensions:"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 352
        },
        "id": "O8ABKNZDEUwQ",
        "outputId": "13b80428-178b-4647-b9a9-ac88ad4b0bf4"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "99.99999999999966\n"
          ]
        }
      ],
      "source": [
        "from sklearn.decomposition import PCA\n",
        "\n",
        "pca_2000 = PCA(n_components=min(data.shape))\n",
        "pca_2000.fit(data)\n",
        "\n",
        "# Visualize the amount of explained variance\n",
        "plt.plot(np.cumsum(pca_2000.explained_variance_ratio_ * 100))\n",
        "plt.grid()\n",
        "plt.xlabel('Number of components')\n",
        "plt.ylabel('Explained variance')\n",
        "plt.show()\n",
        "\n",
        "# Print the amount of explained variance  \n",
        "print(np.cumsum(pca_2000.explained_variance_ratio_ * 100)[-1])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 40,
      "metadata": {
        "id": "4oWdWz2S4kKL"
      },
      "outputs": [],
      "source": [
        "# Do PCA for dimensionality reduction\n",
        "def do_pca(clf, X, y, num_components, label, include_plot=True):\n",
        "\n",
        "    results = []\n",
        "\n",
        "    # Add PCA as a third step of a pipeline\n",
        "    clf.steps.insert(2,['pca', PCA()])\n",
        "\n",
        "    for i in num_components:\n",
        "        # Set the number of features for PCA to 'i'\n",
        "        clf.set_params(pca__n_components=i)\n",
        "        \n",
        "        # Cross-validate and get the accuracy\n",
        "        cv = cross_validate(clf, X, y, cv=5)\n",
        "        scores = cv['test_score']\n",
        "        acc = np.mean(scores)\n",
        "        results.append(acc)\n",
        "    \n",
        "    # Remove from the pipeline\n",
        "    clf.steps.pop(0)\n",
        "\n",
        "    # Show the results\n",
        "    if include_plot:\n",
        "        plt.plot(num_components, results, label=label)\n",
        "        plt.legend(loc='upper right')"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Number of components for pca to be tried\n",
        "num_components = [100, 500, 1000, 1500, 2000, 2700]\n",
        "\n",
        "plt.title('PCA')\n",
        "plt.xlabel('Number of components selected')\n",
        "plt.ylabel('Accuracy')\n",
        "\n",
        "for label, clf in classifiers.items():\n",
        "    do_pca(clf, data, classes, num_components, label)\n",
        "\n",
        "plt.savefig(\"pca.png\", bbox_inches = 'tight')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 350
        },
        "id": "95bMlRkZQEJJ",
        "outputId": "3cbcae6b-219b-4b9e-9475-9ddd0daa1ac5"
      },
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nCz7lFB8ZyoU"
      },
      "source": [
        "## Hyperparameter Tuning"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Select chosen number of features from the dataset\n",
        "feat_idx = clf_results_anova['svm_rbf'][2304]\n",
        "data_reduced_feats = data.iloc[:,feat_idx]\n",
        "data_reduced_feats.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w2D1je7phMBY",
        "outputId": "b567094e-ead1-4790-b885-638630e6999a"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3400, 2304)"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "id": "yJ2FU0UAekTo",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "06031511-24dc-4812-95bc-043c2e5953d9"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "best score: 0.76, parameters: {'clf__C': 10, 'clf__gamma': 0.001, 'clf__kernel': 'rbf'}\n"
          ]
        }
      ],
      "source": [
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "# Create an instance of a classifier \n",
        "clf = Pipeline([('std_scaler', std_scaler), \n",
        "                ('minmax_scaler', minmax_scaler), \n",
        "                ('clf', SVC())])\n",
        "\n",
        "# Create a grid of hyperparameters to be tried\n",
        "svm_param_grid = {'clf__C': [0.01, 0.1, 1, 10],\n",
        "                  'clf__gamma': ['scale', 0.001, 0.01, 0.1],\n",
        "                  'clf__kernel': ['rbf']}  \n",
        "\n",
        "# Find the optimal parameters using 5-fold cv\n",
        "grid_search = GridSearchCV(clf, svm_param_grid, scoring='accuracy', cv=5)\n",
        "search = grid_search.fit(data_reduced_feats, classes) \n",
        "print('best score: {}, parameters: {}'.format(search.best_score_, search.best_params_))"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Apply confidence scores to see how performance changes when some observations are assigned higher weights\n",
        "grid_search_conf = GridSearchCV(clf, svm_param_grid, scoring='accuracy', cv=5)\n",
        "kwargs = {clf.steps[-1][0] + '__sample_weight': confidences*2} # add confidences as sample weight parameter for the SVM\n",
        "search_conf = grid_search_conf.fit(data_reduced_feats, classes, **kwargs) \n",
        "print('best score: {}, parameters: {}'.format(search_conf.best_score_, search_conf.best_params_))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cTNU2xceoyIk",
        "outputId": "142d6b1a-c87c-4eb7-a349-b74c16c44224"
      },
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "best score: 0.7602941176470588, parameters: {'clf__C': 1, 'clf__gamma': 'scale', 'clf__kernel': 'rbf'}\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Evaluation"
      ],
      "metadata": {
        "id": "8Y1P5LPPbKIB"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Show how validation accuracy changes for each combination of hyperparameters\n",
        "scores = search.cv_results_['mean_test_score']\n",
        "\n",
        "plt.xticks(np.arange(16))\n",
        "plt.xlabel('Hyperparameter set')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(scores, label='without conf')\n",
        "\n",
        "# Compare with resulting scores when confidences are used\n",
        "scores_conf = search_conf.cv_results_['mean_test_score']\n",
        "plt.plot(scores_conf, label='with conf')\n",
        "plt.legend()\n",
        "\n",
        "plt.savefig(\"hyperparams.png\", bbox_inches = 'tight')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 334
        },
        "id": "o4fcjTO3kTTe",
        "outputId": "424dad26-3695-48f9-e95f-959bda858d2b"
      },
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Exporting Test Predictions"
      ],
      "metadata": {
        "id": "IMzpTxJBPVNR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Print the number of missing values\n",
        "test_data.isna().sum()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ndxTtoDapg2u",
        "outputId": "fbdbd52a-b2bd-4bdc-b858-3d02b2de7683"
      },
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "CNNs        384\n",
              "CNNs.1      414\n",
              "CNNs.2      390\n",
              "CNNs.3      423\n",
              "CNNs.4      397\n",
              "           ... \n",
              "GIST.507    399\n",
              "GIST.508    437\n",
              "GIST.509    379\n",
              "GIST.510    399\n",
              "GIST.511    375\n",
              "Length: 4608, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Impute missing values with column means\n",
        "test_data = test_data.fillna(value=test_data.mean())\n",
        "\n",
        "# Check if any column has missing values\n",
        "num_nan = test_data.isna().sum().values\n",
        "len(num_nan[num_nan != 0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jf5L3Oy-pic-",
        "outputId": "805ff34a-2711-40fa-97ce-4adc0ba65b5a"
      },
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0"
            ]
          },
          "metadata": {},
          "execution_count": 43
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "test_data.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 299
        },
        "id": "CAqrwZXarntv",
        "outputId": "c90fd2fb-04dd-440b-b52c-36d319f426a7"
      },
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "       CNNs    CNNs.1   CNNs.2    CNNs.3    CNNs.4    CNNs.5   CNNs.6  \\\n",
              "0  0.787990  0.307314  2.02460  0.000000  0.000000  3.295500  0.00000   \n",
              "1  0.000000  0.722650  0.23718  0.158560  0.205002  0.235777  0.00000   \n",
              "2  0.000000  0.307314  0.00000  0.000000  0.648270  0.480990  0.00000   \n",
              "3  0.044644  0.000000  0.23718  0.712940  0.000000  0.235777  0.14967   \n",
              "4  0.848170  0.307314  0.00000  0.304779  0.000000  0.000000  0.25626   \n",
              "\n",
              "    CNNs.7    CNNs.8    CNNs.9  ...  GIST.502  GIST.503  GIST.504  GIST.505  \\\n",
              "0  0.12965  0.000000  0.165616  ...  0.021705  0.034861  0.045078  0.034300   \n",
              "1  0.00000  0.185367  0.000000  ...  0.004348  0.007177  0.014610  0.034300   \n",
              "2  0.71721  0.705410  0.000000  ...  0.026724  0.027204  0.021123  0.027027   \n",
              "3  0.00000  0.000000  0.000000  ...  0.026242  0.029739  0.025212  0.034300   \n",
              "4  0.00000  0.798860  0.000000  ...  0.003493  0.002661  0.032333  0.069194   \n",
              "\n",
              "   GIST.506  GIST.507  GIST.508  GIST.509  GIST.510  GIST.511  \n",
              "0  0.030645  0.031250  0.026715  0.007441  0.028426  0.024078  \n",
              "1  0.007715  0.014736  0.009790  0.036324  0.026978  0.024078  \n",
              "2  0.029328  0.031330  0.030306  0.013995  0.033780  0.013178  \n",
              "3  0.021912  0.020740  0.024278  0.001928  0.015194  0.023487  \n",
              "4  0.040851  0.006043  0.030306  0.035340  0.024834  0.076113  \n",
              "\n",
              "[5 rows x 4608 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-520e04de-4e73-4831-9e53-8159f02def19\">\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>CNNs</th>\n",
              "      <th>CNNs.1</th>\n",
              "      <th>CNNs.2</th>\n",
              "      <th>CNNs.3</th>\n",
              "      <th>CNNs.4</th>\n",
              "      <th>CNNs.5</th>\n",
              "      <th>CNNs.6</th>\n",
              "      <th>CNNs.7</th>\n",
              "      <th>CNNs.8</th>\n",
              "      <th>CNNs.9</th>\n",
              "      <th>...</th>\n",
              "      <th>GIST.502</th>\n",
              "      <th>GIST.503</th>\n",
              "      <th>GIST.504</th>\n",
              "      <th>GIST.505</th>\n",
              "      <th>GIST.506</th>\n",
              "      <th>GIST.507</th>\n",
              "      <th>GIST.508</th>\n",
              "      <th>GIST.509</th>\n",
              "      <th>GIST.510</th>\n",
              "      <th>GIST.511</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0.787990</td>\n",
              "      <td>0.307314</td>\n",
              "      <td>2.02460</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>3.295500</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.12965</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.165616</td>\n",
              "      <td>...</td>\n",
              "      <td>0.021705</td>\n",
              "      <td>0.034861</td>\n",
              "      <td>0.045078</td>\n",
              "      <td>0.034300</td>\n",
              "      <td>0.030645</td>\n",
              "      <td>0.031250</td>\n",
              "      <td>0.026715</td>\n",
              "      <td>0.007441</td>\n",
              "      <td>0.028426</td>\n",
              "      <td>0.024078</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.722650</td>\n",
              "      <td>0.23718</td>\n",
              "      <td>0.158560</td>\n",
              "      <td>0.205002</td>\n",
              "      <td>0.235777</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.185367</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.004348</td>\n",
              "      <td>0.007177</td>\n",
              "      <td>0.014610</td>\n",
              "      <td>0.034300</td>\n",
              "      <td>0.007715</td>\n",
              "      <td>0.014736</td>\n",
              "      <td>0.009790</td>\n",
              "      <td>0.036324</td>\n",
              "      <td>0.026978</td>\n",
              "      <td>0.024078</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.307314</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.648270</td>\n",
              "      <td>0.480990</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.71721</td>\n",
              "      <td>0.705410</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.026724</td>\n",
              "      <td>0.027204</td>\n",
              "      <td>0.021123</td>\n",
              "      <td>0.027027</td>\n",
              "      <td>0.029328</td>\n",
              "      <td>0.031330</td>\n",
              "      <td>0.030306</td>\n",
              "      <td>0.013995</td>\n",
              "      <td>0.033780</td>\n",
              "      <td>0.013178</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.044644</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.23718</td>\n",
              "      <td>0.712940</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.235777</td>\n",
              "      <td>0.14967</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.026242</td>\n",
              "      <td>0.029739</td>\n",
              "      <td>0.025212</td>\n",
              "      <td>0.034300</td>\n",
              "      <td>0.021912</td>\n",
              "      <td>0.020740</td>\n",
              "      <td>0.024278</td>\n",
              "      <td>0.001928</td>\n",
              "      <td>0.015194</td>\n",
              "      <td>0.023487</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.848170</td>\n",
              "      <td>0.307314</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.304779</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.25626</td>\n",
              "      <td>0.00000</td>\n",
              "      <td>0.798860</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>...</td>\n",
              "      <td>0.003493</td>\n",
              "      <td>0.002661</td>\n",
              "      <td>0.032333</td>\n",
              "      <td>0.069194</td>\n",
              "      <td>0.040851</td>\n",
              "      <td>0.006043</td>\n",
              "      <td>0.030306</td>\n",
              "      <td>0.035340</td>\n",
              "      <td>0.024834</td>\n",
              "      <td>0.076113</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 4608 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-520e04de-4e73-4831-9e53-8159f02def19')\"\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-520e04de-4e73-4831-9e53-8159f02def19 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-520e04de-4e73-4831-9e53-8159f02def19');\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",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 44
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Reduce the number of features in the testing dataset\n",
        "feat_idx = clf_results_anova['svm_rbf'][2304]\n",
        "test_data_reduced_feats = test_data.iloc[:,feat_idx]\n",
        "test_data_reduced_feats.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cvI-npCrplIk",
        "outputId": "9c19955d-8e6e-4386-b7c8-c86447635e96"
      },
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(2000, 2304)"
            ]
          },
          "metadata": {},
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Create an instance of a classifier \n",
        "clf = Pipeline([('std_scaler', std_scaler), \n",
        "                ('minmax_scaler', minmax_scaler), \n",
        "                ('clf', SVC())])\n",
        "\n",
        "# Set optimal parameters\n",
        "best_params = search_conf.best_params_\n",
        "clf.set_params(**best_params)\n",
        "\n",
        "# Add confidences as sample weight parameter for the classifier\n",
        "kwargs = {clf.steps[-1][0] + '__sample_weight': confidences*2}\n",
        "\n",
        "# Train the classifier\n",
        "clf.fit(data_reduced_feats, classes, **kwargs)\n",
        "\n",
        "# Predict class labels for the test dataset\n",
        "predictions = clf.predict(test_data_reduced_feats)\n",
        "\n",
        "# Export predictions\n",
        "predictions = pd.DataFrame(predictions, columns=['label']) # convert to DataFrame\n",
        "predictions.to_csv('predictions.csv', index=False)"
      ],
      "metadata": {
        "id": "wVaM9HqxPXXb"
      },
      "execution_count": 62,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "name": "ReportML.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}