Thyroid_Disease_Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Thyroid_disease_classification_Forza.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gsqC_gTs439C"
      },
      "source": [
        "#### Load dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mwk5v1TnRMUI"
      },
      "source": [
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "import numpy as np"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BGEvSCG8tTxP"
      },
      "source": [
        "# list of column names\n",
        "columns = [\"Age\", \"Sex\", \"On Thyroxine\", \"Query on Thyroxine\", \n",
        "           \"On Antithyroid Medication\", \"Sick\", \"Pregnant\", \n",
        "           \"Thyroid Surgery\", \"I131 Treatment\", \"Query Hypothyroid\", \n",
        "           \"Query Hyperthyroid\", \"Lithium\", \"Goitre\", \"Tumor\", \n",
        "           \"Hypopituitary\", \"Psych\", \"TSH Measured\", \"TSH\", \"T3 Measured\", \n",
        "           \"T3\", \"TT4 Measured\", \"TT4\", \"T4U Measured\", \"T4U\", \n",
        "           \"FTI Measured\", \"FTI\", \"TBG Measured\", \"TBG\", \"Referral Source\", \"Category\"]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0LDKIzCQSNLo"
      },
      "source": [
        "# reading dataset\n",
        "data=pd.read_csv('/content/data.csv',na_values='?',header=None, names=columns)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_jCuHxO_4HZL"
      },
      "source": [
        "# extract category names from target column values \n",
        "data.iloc[:,-1] = data.iloc[:,-1].str.extract('([a-zA-Z0-9\\s]*)')[0].str.lower()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z1XM3AQCEj8a"
      },
      "source": [
        "# remove duplicate entries\n",
        "data=data.drop_duplicates()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r90d658vTbMZ"
      },
      "source": [
        "# shuffling dataset\n",
        "data = data.sample(frac = 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JwEz4HOfwuOo"
      },
      "source": [
        "# splitting features and target columns\n",
        "X = data.iloc[:,:-1] \n",
        "y = data.iloc[:,-1]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-xZT0ndkxCkM"
      },
      "source": [
        "# generalise categories\n",
        "y.replace(to_replace=['compensated hypothyroid','primary hypothyroid','secondary hypothyroid'], value='hypothyroid', inplace=True)\n",
        "y.replace(to_replace=['goitre','hyperthyroid','t3 toxic'], value='hyperthyroid', inplace=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 478
        },
        "id": "TbXAbe_6hFIk",
        "outputId": "81894d58-3a2e-45dc-8037-8b4c7ac9397f"
      },
      "source": [
        "X"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Age</th>\n",
              "      <th>Sex</th>\n",
              "      <th>On Thyroxine</th>\n",
              "      <th>Query on Thyroxine</th>\n",
              "      <th>On Antithyroid Medication</th>\n",
              "      <th>Sick</th>\n",
              "      <th>Pregnant</th>\n",
              "      <th>Thyroid Surgery</th>\n",
              "      <th>I131 Treatment</th>\n",
              "      <th>Query Hypothyroid</th>\n",
              "      <th>Query Hyperthyroid</th>\n",
              "      <th>Lithium</th>\n",
              "      <th>Goitre</th>\n",
              "      <th>Tumor</th>\n",
              "      <th>Hypopituitary</th>\n",
              "      <th>Psych</th>\n",
              "      <th>TSH Measured</th>\n",
              "      <th>TSH</th>\n",
              "      <th>T3 Measured</th>\n",
              "      <th>T3</th>\n",
              "      <th>TT4 Measured</th>\n",
              "      <th>TT4</th>\n",
              "      <th>T4U Measured</th>\n",
              "      <th>T4U</th>\n",
              "      <th>FTI Measured</th>\n",
              "      <th>FTI</th>\n",
              "      <th>TBG Measured</th>\n",
              "      <th>TBG</th>\n",
              "      <th>Referral Source</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1227</th>\n",
              "      <td>31.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>1.900</td>\n",
              "      <td>t</td>\n",
              "      <td>2.7</td>\n",
              "      <td>t</td>\n",
              "      <td>99.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.00</td>\n",
              "      <td>t</td>\n",
              "      <td>99.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>STMW</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2771</th>\n",
              "      <td>71.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>1.500</td>\n",
              "      <td>t</td>\n",
              "      <td>2.0</td>\n",
              "      <td>t</td>\n",
              "      <td>96.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.05</td>\n",
              "      <td>t</td>\n",
              "      <td>92.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>SVI</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2328</th>\n",
              "      <td>74.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>1.300</td>\n",
              "      <td>t</td>\n",
              "      <td>1.9</td>\n",
              "      <td>t</td>\n",
              "      <td>99.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.12</td>\n",
              "      <td>t</td>\n",
              "      <td>88.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>other</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3840</th>\n",
              "      <td>72.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>2.300</td>\n",
              "      <td>t</td>\n",
              "      <td>0.8</td>\n",
              "      <td>t</td>\n",
              "      <td>90.0</td>\n",
              "      <td>t</td>\n",
              "      <td>0.80</td>\n",
              "      <td>t</td>\n",
              "      <td>112.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>SVI</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1715</th>\n",
              "      <td>15.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>2.600</td>\n",
              "      <td>t</td>\n",
              "      <td>2.4</td>\n",
              "      <td>t</td>\n",
              "      <td>100.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>other</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",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3584</th>\n",
              "      <td>59.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>7.100</td>\n",
              "      <td>t</td>\n",
              "      <td>1.9</td>\n",
              "      <td>t</td>\n",
              "      <td>98.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.11</td>\n",
              "      <td>t</td>\n",
              "      <td>88.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>SVHD</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2390</th>\n",
              "      <td>34.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>4.100</td>\n",
              "      <td>t</td>\n",
              "      <td>4.0</td>\n",
              "      <td>t</td>\n",
              "      <td>159.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.62</td>\n",
              "      <td>t</td>\n",
              "      <td>98.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>STMW</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>90</th>\n",
              "      <td>63.0</td>\n",
              "      <td>F</td>\n",
              "      <td>t</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>0.025</td>\n",
              "      <td>t</td>\n",
              "      <td>1.6</td>\n",
              "      <td>t</td>\n",
              "      <td>124.0</td>\n",
              "      <td>t</td>\n",
              "      <td>0.89</td>\n",
              "      <td>t</td>\n",
              "      <td>139.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>other</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1246</th>\n",
              "      <td>62.0</td>\n",
              "      <td>F</td>\n",
              "      <td>t</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>2.500</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>t</td>\n",
              "      <td>126.0</td>\n",
              "      <td>t</td>\n",
              "      <td>1.24</td>\n",
              "      <td>t</td>\n",
              "      <td>101.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>other</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2316</th>\n",
              "      <td>51.0</td>\n",
              "      <td>F</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>f</td>\n",
              "      <td>t</td>\n",
              "      <td>24.000</td>\n",
              "      <td>t</td>\n",
              "      <td>0.4</td>\n",
              "      <td>t</td>\n",
              "      <td>3.0</td>\n",
              "      <td>t</td>\n",
              "      <td>0.99</td>\n",
              "      <td>t</td>\n",
              "      <td>3.0</td>\n",
              "      <td>f</td>\n",
              "      <td>NaN</td>\n",
              "      <td>other</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>3221 rows × 29 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "       Age  Sex On Thyroxine  ... TBG Measured TBG Referral Source\n",
              "1227  31.0  NaN            f  ...            f NaN            STMW\n",
              "2771  71.0    F            f  ...            f NaN             SVI\n",
              "2328  74.0    F            f  ...            f NaN           other\n",
              "3840  72.0    F            f  ...            f NaN             SVI\n",
              "1715  15.0    F            f  ...            f NaN           other\n",
              "...    ...  ...          ...  ...          ...  ..             ...\n",
              "3584  59.0    F            f  ...            f NaN            SVHD\n",
              "2390  34.0    F            f  ...            f NaN            STMW\n",
              "90    63.0    F            t  ...            f NaN           other\n",
              "1246  62.0    F            t  ...            f NaN           other\n",
              "2316  51.0    F            f  ...            f NaN           other\n",
              "\n",
              "[3221 rows x 29 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "daQQrNCM5EKj"
      },
      "source": [
        "#### Exploratory Analysis of the Dataset \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GBCVUWsCBSC2",
        "outputId": "ce2dee48-b71c-4157-c981-481b30d2607d"
      },
      "source": [
        "# print the names of the features\n",
        "print(\"Features: \", list(X.columns))\n",
        "\n",
        "# print the label types of target column\n",
        "print(\"Labels: \", set(y))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Features:  ['Age', 'Sex', 'On Thyroxine', 'Query on Thyroxine', 'On Antithyroid Medication', 'Sick', 'Pregnant', 'Thyroid Surgery', 'I131 Treatment', 'Query Hypothyroid', 'Query Hyperthyroid', 'Lithium', 'Goitre', 'Tumor', 'Hypopituitary', 'Psych', 'TSH Measured', 'TSH', 'T3 Measured', 'T3', 'TT4 Measured', 'TT4', 'T4U Measured', 'T4U', 'FTI Measured', 'FTI', 'TBG Measured', 'TBG', 'Referral Source']\n",
            "Labels:  {'sick', 'hyperthyroid', 'negative', 'hypothyroid'}\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "85V08SxpT2R2",
        "outputId": "0bc36280-426d-49c0-c3ae-e47ba42df0ab"
      },
      "source": [
        "print('Shape:',data.shape)\n",
        "#Dataset comprises of 3221 observations and 30 characteristics (29 independant and 1 dependant)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shape: (3221, 30)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5Ai5I5rBUe9g",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "07a20620-c1d9-4d60-f448-5ca7d74a6080"
      },
      "source": [
        "data.info()\n",
        "# float -> numerical columns, object -> nominal/categorical columns \n",
        "# Data has 7 columns of float dtype, 23 of object(string) dtype\n",
        "# 27 column has too many missing values"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "Int64Index: 3221 entries, 1227 to 2316\n",
            "Data columns (total 30 columns):\n",
            " #   Column                     Non-Null Count  Dtype  \n",
            "---  ------                     --------------  -----  \n",
            " 0   Age                        3220 non-null   float64\n",
            " 1   Sex                        3094 non-null   object \n",
            " 2   On Thyroxine               3221 non-null   object \n",
            " 3   Query on Thyroxine         3221 non-null   object \n",
            " 4   On Antithyroid Medication  3221 non-null   object \n",
            " 5   Sick                       3221 non-null   object \n",
            " 6   Pregnant                   3221 non-null   object \n",
            " 7   Thyroid Surgery            3221 non-null   object \n",
            " 8   I131 Treatment             3221 non-null   object \n",
            " 9   Query Hypothyroid          3221 non-null   object \n",
            " 10  Query Hyperthyroid         3221 non-null   object \n",
            " 11  Lithium                    3221 non-null   object \n",
            " 12  Goitre                     3221 non-null   object \n",
            " 13  Tumor                      3221 non-null   object \n",
            " 14  Hypopituitary              3221 non-null   object \n",
            " 15  Psych                      3221 non-null   object \n",
            " 16  TSH Measured               3221 non-null   object \n",
            " 17  TSH                        2974 non-null   float64\n",
            " 18  T3 Measured                3221 non-null   object \n",
            " 19  T3                         2632 non-null   float64\n",
            " 20  TT4 Measured               3221 non-null   object \n",
            " 21  TT4                        3079 non-null   float64\n",
            " 22  T4U Measured               3221 non-null   object \n",
            " 23  T4U                        2945 non-null   float64\n",
            " 24  FTI Measured               3221 non-null   object \n",
            " 25  FTI                        2947 non-null   float64\n",
            " 26  TBG Measured               3221 non-null   object \n",
            " 27  TBG                        0 non-null      float64\n",
            " 28  Referral Source            3221 non-null   object \n",
            " 29  Category                   3221 non-null   object \n",
            "dtypes: float64(7), object(23)\n",
            "memory usage: 780.1+ KB\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "ZPAuKqGGU5ch",
        "outputId": "3dac8d7e-62f3-4314-96cd-a9c31cdaafb3"
      },
      "source": [
        "data.describe()\n",
        "### There is a large difference between 75th %tile and max value of 'Age' and 'TSH' which implies that the outliers have notably larger values"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Age</th>\n",
              "      <th>TSH</th>\n",
              "      <th>T3</th>\n",
              "      <th>TT4</th>\n",
              "      <th>T4U</th>\n",
              "      <th>FTI</th>\n",
              "      <th>TBG</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>3220.000000</td>\n",
              "      <td>2974.000000</td>\n",
              "      <td>2632.000000</td>\n",
              "      <td>3079.000000</td>\n",
              "      <td>2945.000000</td>\n",
              "      <td>2947.000000</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>52.532609</td>\n",
              "      <td>6.722840</td>\n",
              "      <td>1.963355</td>\n",
              "      <td>107.807665</td>\n",
              "      <td>0.989937</td>\n",
              "      <td>110.657550</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>20.381326</td>\n",
              "      <td>27.585818</td>\n",
              "      <td>0.928773</td>\n",
              "      <td>38.941172</td>\n",
              "      <td>0.194417</td>\n",
              "      <td>37.578197</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.005000</td>\n",
              "      <td>0.050000</td>\n",
              "      <td>2.000000</td>\n",
              "      <td>0.310000</td>\n",
              "      <td>2.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>37.000000</td>\n",
              "      <td>0.450000</td>\n",
              "      <td>1.500000</td>\n",
              "      <td>86.000000</td>\n",
              "      <td>0.870000</td>\n",
              "      <td>92.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>55.000000</td>\n",
              "      <td>1.500000</td>\n",
              "      <td>1.900000</td>\n",
              "      <td>102.000000</td>\n",
              "      <td>0.970000</td>\n",
              "      <td>106.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>68.000000</td>\n",
              "      <td>3.300000</td>\n",
              "      <td>2.300000</td>\n",
              "      <td>125.000000</td>\n",
              "      <td>1.080000</td>\n",
              "      <td>125.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>455.000000</td>\n",
              "      <td>478.000000</td>\n",
              "      <td>10.600000</td>\n",
              "      <td>430.000000</td>\n",
              "      <td>2.120000</td>\n",
              "      <td>395.000000</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               Age          TSH           T3  ...          T4U          FTI  TBG\n",
              "count  3220.000000  2974.000000  2632.000000  ...  2945.000000  2947.000000  0.0\n",
              "mean     52.532609     6.722840     1.963355  ...     0.989937   110.657550  NaN\n",
              "std      20.381326    27.585818     0.928773  ...     0.194417    37.578197  NaN\n",
              "min       1.000000     0.005000     0.050000  ...     0.310000     2.000000  NaN\n",
              "25%      37.000000     0.450000     1.500000  ...     0.870000    92.000000  NaN\n",
              "50%      55.000000     1.500000     1.900000  ...     0.970000   106.000000  NaN\n",
              "75%      68.000000     3.300000     2.300000  ...     1.080000   125.000000  NaN\n",
              "max     455.000000   478.000000    10.600000  ...     2.120000   395.000000  NaN\n",
              "\n",
              "[8 rows x 7 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YbCqKLQM5NNn"
      },
      "source": [
        "#####\tPerform univariate analysis by plotting various charts \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 597
        },
        "id": "x9Ru1ARvVsQv",
        "outputId": "28f5ac96-33b6-4454-b1ef-801cd15048fa"
      },
      "source": [
        "# histograms\n",
        "data.hist(alpha=0.5,figsize=(8,8))\n",
        "plt.tight_layout(pad=0.5)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 9 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 381
        },
        "id": "eH3bnJT9WHCe",
        "outputId": "f8797c45-731f-47c6-9bfc-ab995d361ef6"
      },
      "source": [
        "# boxplots\n",
        "# box plots to view distribution of values across their range and identify which columns have outliers\n",
        "numeric_col = ['Age','TSH','T3','TT4','T4U','FTI']\n",
        "\n",
        "fig, axs = plt.subplots(ncols=6, nrows=1, figsize=(10, 5))\n",
        "axs = axs.flatten()\n",
        "i = 0\n",
        "for c in numeric_col:\n",
        "    sns.boxplot(y=c, data=data, ax=axs[i])\n",
        "    i += 1\n",
        "plt.tight_layout(pad=0.5, w_pad=0.5, h_pad=5.0)\n",
        "# all numeric columns have a lot of outliers"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x360 with 6 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G7v4jsXr5WPE"
      },
      "source": [
        "##### Perform multivariate analysis \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kTuz7xdqXEy6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 613
        },
        "outputId": "a71ae6f9-b063-41e8-cf71-934e827822dc"
      },
      "source": [
        "# correlation matrix \n",
        "plt.figure(figsize=(10,10))\n",
        "sns.heatmap(data=X.corr(),cmap='Blues',annot=True,linewidths=0.2)\n",
        "# TT4 and FTI columns are highly correlated"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f2c07457250>"
            ]
          },
          "metadata": {},
          "execution_count": 29
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bk4hTdOS5g_K"
      },
      "source": [
        "#### Missing Value Imputation and removing any undesirable feature from the dataset."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kGMz81rEYTq4"
      },
      "source": [
        "# {} column is unique to almost each row and does not contribute anything to the classification model\n",
        "# {} column has too many null values\n",
        "X = X.drop(['TBG','TBG Measured'],axis=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YFwoUqBDX5ze"
      },
      "source": [
        "# Imputing missing values using most_frequent strategy\n",
        "from sklearn.impute import SimpleImputer\n",
        "imp = SimpleImputer(missing_values=np.NaN, strategy='most_frequent')\n",
        "X = pd.DataFrame(imp.fit_transform(X),columns=X.columns)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pwFxuaF9l1AV"
      },
      "source": [
        "#### Encoding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DOq3L-u8gK6V"
      },
      "source": [
        "# function to one hot encode given column in given dataset\n",
        "def ohe(data,i):\n",
        "  # store one hot encoded columns in temp\n",
        "  temp = pd.get_dummies(data[i],prefix=i)\n",
        "  # store index of given column in t\n",
        "  t = data.columns.get_loc(i)\n",
        "  # drop given column \n",
        "  data = data.drop(labels=i,axis=1)\n",
        "  # insert one hot encoded columns in index of the given column\n",
        "  for i in range(temp.shape[1]-1,-1,-1):\n",
        "    data.insert(t, temp.columns[i], temp.iloc[:,i],False)\n",
        "  return data\n",
        "\n",
        "# function to label encode given column in given dataset\n",
        "def le(data,i):\n",
        "  from sklearn.preprocessing import LabelEncoder \n",
        "  # standardize the string (remove apostrophes [if they exist] and lowercase all alphabets)\n",
        "  #data[i] = data[i].str.extract('\\'?([a-zA-Z\\s\\-\\[\\]\\/]*)\\'?')[0].str.lower()\n",
        "  # label encode given column\n",
        "  data[i] = pd.DataFrame(LabelEncoder().fit_transform(data[i]),columns=[i])\n",
        "  return data\n",
        "\n",
        "# lists to store all column names which need to get one hot encoded or label encoded\n",
        "# required for encoding of features sub-step of Accuracy Improvement Step \n",
        "#ohe_names=[] \n",
        "#le_names=[]\n",
        "\n",
        "# iterate through all the columns in data\n",
        "for i in X.columns:\n",
        "  # if column values are not numeric\n",
        "  if type(X[i][0]) not in (np.int64,np.float64):\n",
        "    # if unique values in column are more than 5 then label encode the column\n",
        "    if len(set(X[i]))>5: \n",
        "      X=le(X,i)\n",
        "      #le_names.append(i) # append all column names which got label encoded\n",
        "    # if unique values in column are less than equal to 5 one hot encode the column\n",
        "    else: \n",
        "      X=ohe(X,i)\n",
        "      #ohe_names.append(i) # append all column names which got one hot encoded\n",
        "\n",
        "#d2 = X.copy() # for later step accuracy improvement by removal of some features"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CCLwMigNXWma"
      },
      "source": [
        "####Remove unwanted features/feature selection \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "202FT5GRXwRt"
      },
      "source": [
        "from sklearn.feature_selection import SelectKBest\n",
        "#apply SelectKBest class to extract top 15 best features using ANOVA f-score to evaluate\n",
        "X = pd.DataFrame(SelectKBest(k=15).fit_transform(X,y))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Dt-qTJNrmBnT"
      },
      "source": [
        "#### Normalization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XYeG_xBUkouu"
      },
      "source": [
        "# Normalising Data\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "X = pd.DataFrame(MinMaxScaler().fit_transform(X),columns=X.columns)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4kzBIbyC52gW"
      },
      "source": [
        "#### Check for the outliers in the columns and treat the outliers if present."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OXTtgd8icYJI"
      },
      "source": [
        "eps=1e-7\n",
        "# do log transformation of columns with outliers till there are no outliers left (should run max twice)\n",
        "while 1:\n",
        "  # checking outliers by using skewness from normal distribution\n",
        "  o_col=[]\n",
        "  skew = X.skew()\n",
        "  for c in X.columns:\n",
        "    if abs(skew[c])>1:\n",
        "      o_col.append(c)\n",
        "  # end the loop if there are no columns with outliers\n",
        "  if not o_col: break\n",
        "  # treating outliers by log transforming all columns with skewness\n",
        "  for c in o_col:\n",
        "      X[c] = np.maximum(np.full(X[c].shape, eps), np.minimum(np.full(X[c].shape, 1-eps), X[c]))\n",
        "      X[c] = np.log10(X[c])\n",
        "  break"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NoAT2_mx58Pm"
      },
      "source": [
        "#### Split the dataset into train and test. \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P4VAWj3HbIJi"
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,y)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mQdatWmVhfbs"
      },
      "source": [
        "#### Construct models "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1x53mngMr0_3"
      },
      "source": [
        "def pred(model,X_train,X_test):\n",
        "  return model.predict(X_train),model.predict(X_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "unnUBSHDusij"
      },
      "source": [
        "y_train_pred_bef=[]\n",
        "y_test_pred_bef=[]\n",
        "\n",
        "y_train_pred_af=[]\n",
        "y_test_pred_af=[]\n",
        "\n",
        "model_bef=[]\n",
        "model_af=[]\n",
        "\n",
        "model_names=['Logistic Regression','Naive Bayes','Decision Tree','Random Forest','SVM','Gradient Boost','Adaboost','Bagging']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Og7JYbWFU68X"
      },
      "source": [
        "##### Logistic Regression "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ILLOAHtuRSGz",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9b6b3894-9f66-4bea-c64e-ce16c357b2c3"
      },
      "source": [
        "# logistic regression and naive bayes have not particularly tuned \n",
        "from sklearn.linear_model import LogisticRegression\n",
        "\n",
        "model = LogisticRegression(max_iter=1000).fit(X_train, y_train) \n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)\n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KgJxk1FBVBmJ"
      },
      "source": [
        "##### Naive Bayes  "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CnLQzusZhmM_"
      },
      "source": [
        "from sklearn.naive_bayes import GaussianNB\n",
        "\n",
        "model = GaussianNB().fit(X_train,y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)\n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "c2dvtNqGVEnY"
      },
      "source": [
        "##### Decision Tree "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9ypGQuatGgvA"
      },
      "source": [
        "from sklearn.tree import DecisionTreeClassifier\n",
        "\n",
        "model = DecisionTreeClassifier().fit(X_train, y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4N-MdFboK8i-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ef0c7b91-11fc-46d4-b78f-98b7161993f3"
      },
      "source": [
        "# removed some non-crucial parameters because the default google colab environment is not designed for compute intensive machine learning workload\n",
        "from sklearn.tree import DecisionTreeClassifier\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "param = {#'criterion'               :['gini','entropy'],\n",
        "         #'splitter'                :['best','random'],\n",
        "         'max_depth'               :[None,2,5],\n",
        "         'min_samples_split'       :[2,3,4],\n",
        "         'min_samples_leaf'        :[1,2,3],\n",
        "         #'min_weight_fraction_leaf':[0,0.1,0.2],\n",
        "         'max_leaf_nodes'          :[None,2,3],\n",
        "         #'min_impurity_decrease'   :[0,0.1,0.2],\n",
        "         'class_weight'            :[None,'balanced'],\n",
        "         #'ccp_alpha'               :[0,0.1,0.2]\n",
        "         }\n",
        "model = DecisionTreeClassifier(random_state=1,max_features='auto')\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=DecisionTreeClassifier(max_features='auto',\n",
              "                                              random_state=1),\n",
              "             param_grid={'class_weight': [None, 'balanced'],\n",
              "                         'max_depth': [None, 2, 5],\n",
              "                         'max_leaf_nodes': [None, 2, 3],\n",
              "                         'min_samples_leaf': [1, 2, 3],\n",
              "                         'min_samples_split': [2, 3, 4]})"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EJDaz5NbK8i-",
        "outputId": "bfaefa86-e568-4817-ce01-9ceaba937877"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best Decision Tree Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best Decision Tree Model: \n",
            "class_weight: None\n",
            "max_depth: None\n",
            "max_leaf_nodes: 2\n",
            "min_samples_leaf: 1\n",
            "min_samples_split: 2\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A98KlaW6jYul"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZKVWnQvvVNSM"
      },
      "source": [
        "##### Random Forest"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oz1HPZLYj9jl"
      },
      "source": [
        "from sklearn.ensemble import RandomForestClassifier\n",
        "\n",
        "model = RandomForestClassifier().fit(X_train,y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FO2vcqul4tXQ",
        "outputId": "bb6926dc-5d05-48a8-cdb4-5a81f4ccdf2c"
      },
      "source": [
        "from sklearn.ensemble import RandomForestClassifier\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "param = {'n_estimators'            :[2,3,5,10],\n",
        "         #'criterion'               :['gini','entropy'],\n",
        "         'max_depth'               :[None,1,2],\n",
        "         'min_samples_split'       :[2,70,0.05], \n",
        "         'min_samples_leaf'        :[0.01,0.02,0.03],\n",
        "         #'min_weight_fraction_leaf':[0,0.1,0.2],\n",
        "         #'max_features'            :[None,'sqrt','log2'],\n",
        "         #'max_leaf_nodes'          :[None,2,3],\n",
        "         'min_impurity_decrease'   :[0,0.1,0.2],\n",
        "         #'bootstrap'               :[True,False],\n",
        "         #'oob_score'               :[True,False], \n",
        "         #'n_jobs'                  :[None,-1,2],\n",
        "         #'random_state'            :[None,1,2],\n",
        "         #'class_weight'            :[None,'balanced'],\n",
        "         #'ccp_alpha'               :[0,0.1,0.2],\n",
        "         #'max_samples'             :[None,25,50]\n",
        "         }\n",
        "model = RandomForestClassifier()\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=RandomForestClassifier(),\n",
              "             param_grid={'max_depth': [None, 1, 2],\n",
              "                         'min_impurity_decrease': [0, 0.1, 0.2],\n",
              "                         'min_samples_leaf': [0.01, 0.02, 0.03],\n",
              "                         'min_samples_split': [2, 70, 0.05],\n",
              "                         'n_estimators': [2, 3, 5, 10]})"
            ]
          },
          "metadata": {},
          "execution_count": 46
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rqro-zd-lJ1G",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4db62a54-a959-44c1-da00-58ff09ec502e"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best Random Forest Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best Random Forest Model: \n",
            "max_depth: None\n",
            "min_impurity_decrease: 0\n",
            "min_samples_leaf: 0.01\n",
            "min_samples_split: 0.05\n",
            "n_estimators: 10\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nO9L9uK0jWie"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RDFbnjaVVQej"
      },
      "source": [
        "##### SVM\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5OjmDVXTkX5Y"
      },
      "source": [
        "from sklearn.svm import SVC\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "model = SVC(class_weight='balanced').fit(X_train, y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2w-mk5JDL4UA",
        "outputId": "7792e175-8279-4881-e884-f103145ceafd"
      },
      "source": [
        "from sklearn.svm import SVC\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "param = {'C'        :[1,2,4,8], \n",
        "         'kernel'   :['poly', 'rbf', 'sigmoid'],\n",
        "         'degree'   :[3,4,5],\n",
        "         'gamma'    :['auto','scale']\n",
        "        }\n",
        "model = SVC(class_weight='balanced')\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=SVC(class_weight='balanced'),\n",
              "             param_grid={'C': [1, 2, 4, 8], 'degree': [3, 4, 5],\n",
              "                         'gamma': ['auto', 'scale'],\n",
              "                         'kernel': ['poly', 'rbf', 'sigmoid']})"
            ]
          },
          "metadata": {},
          "execution_count": 50
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9b-ADgitPFyu",
        "outputId": "8618552e-7cd8-40d0-bc23-55ccccfe0836"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best SVC Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best SVC Model: \n",
            "C: 2\n",
            "degree: 5\n",
            "gamma: auto\n",
            "kernel: poly\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rwhvEDUDjT2H"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3eXsgh1MVS6F"
      },
      "source": [
        "##### Gradient Boost"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uPpDz_CHnkde"
      },
      "source": [
        "from sklearn.ensemble import GradientBoostingClassifier\n",
        "\n",
        "model = GradientBoostingClassifier().fit(X_train, y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UJkNFOswSAji",
        "outputId": "13468242-bab6-4fa7-c492-7e534839020e"
      },
      "source": [
        "from sklearn.ensemble import GradientBoostingClassifier\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "param = {#'loss'                     :['deviance','exponential'], \n",
        "         'learning_rate'            :[0.1],\n",
        "         'n_estimators'             :[100],\n",
        "         #'subsample'                :[],\n",
        "         'min_samples_split'        :[2,70,0.05], \n",
        "         'min_samples_leaf'         :[0.01,0.02,0.03],\n",
        "         #'min_weight_fraction_leaf' :[],\n",
        "         'max_depth'                :[None, 1, 2],\n",
        "         #'min_impurity_decrease'    :[],\n",
        "         'random_state'             :[1]\n",
        "        }\n",
        "model = GradientBoostingClassifier()\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=GradientBoostingClassifier(),\n",
              "             param_grid={'learning_rate': [0.1], 'max_depth': [None, 1, 2],\n",
              "                         'min_samples_leaf': [0.01, 0.02, 0.03],\n",
              "                         'min_samples_split': [2, 70, 0.05],\n",
              "                         'n_estimators': [100], 'random_state': [1]})"
            ]
          },
          "metadata": {},
          "execution_count": 54
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GLrlrVdbSAji",
        "outputId": "3d9c93c9-4f0c-4489-d060-56f23df2726e"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best Gradient Boost Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best Gradient Boost Model: \n",
            "learning_rate: 0.1\n",
            "max_depth: 1\n",
            "min_samples_leaf: 0.03\n",
            "min_samples_split: 2\n",
            "n_estimators: 100\n",
            "random_state: 1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "azCQzHikjRII"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "beTz0HS8VVpW"
      },
      "source": [
        "##### Adaboost"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HCYuM0SdntvC"
      },
      "source": [
        "from sklearn.ensemble import AdaBoostClassifier\n",
        "\n",
        "model = AdaBoostClassifier().fit(X_train, y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PsMWBwuOT-ep",
        "outputId": "cc6e51bd-5180-408f-a762-3436453352d3"
      },
      "source": [
        "from sklearn.ensemble import AdaBoostClassifier\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.naive_bayes import GaussianNB\n",
        "\n",
        "param = {'base_estimator'           :[None,LogisticRegression(),GaussianNB()],\n",
        "         'n_estimators'             :[5,10,15,50],\n",
        "         'learning_rate'            :[1.0],\n",
        "         'random_state'             :[1]\n",
        "        }\n",
        "model = AdaBoostClassifier()\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=AdaBoostClassifier(),\n",
              "             param_grid={'base_estimator': [None, LogisticRegression(),\n",
              "                                            GaussianNB()],\n",
              "                         'learning_rate': [1.0],\n",
              "                         'n_estimators': [5, 10, 15, 50], 'random_state': [1]})"
            ]
          },
          "metadata": {},
          "execution_count": 58
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mfMBHDlDT-eq",
        "outputId": "a7f4dba6-75e1-4e6e-8d2f-1cb9f897bbc1"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best AdaBoost Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best AdaBoost Model: \n",
            "base_estimator: LogisticRegression()\n",
            "learning_rate: 1.0\n",
            "n_estimators: 10\n",
            "random_state: 1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Lpo2Zbu9jNFB"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Km9MP7AMVYuG"
      },
      "source": [
        "##### Bagging"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vqXMLwIjgnJn"
      },
      "source": [
        "from sklearn.ensemble import BaggingClassifier\n",
        "\n",
        "model=BaggingClassifier().fit(X_train,y_train)\n",
        "train_pred,test_pred = pred(model,X_train,X_test) \n",
        "\n",
        "y_train_pred_bef.append(train_pred)\n",
        "y_test_pred_bef.append(test_pred)\n",
        "model_bef.append(model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cXzO4e5mLYkd",
        "outputId": "d521af9e-d04c-428f-d488-2568a1bbce68"
      },
      "source": [
        "from sklearn.ensemble import BaggingClassifier\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.naive_bayes import GaussianNB\n",
        "param = {'base_estimator'          :[None,LogisticRegression(max_iter=1000),GaussianNB()],\n",
        "         'n_estimators'            :[5,10,15],\n",
        "         'max_samples'             :[1.0,100,0.5],\n",
        "         #'max_features'            :[1.0,5,0.5],\n",
        "         #'bootstrap'               :[True,False],\n",
        "         #'bootstrap_features'      :[True,False],\n",
        "         #'oob_score'               :[True,False], \n",
        "         #'n_jobs'                  :[None,-1,2],\n",
        "         'random_state'            :[1],\n",
        "         }\n",
        "model = BaggingClassifier()\n",
        "clf = GridSearchCV(model,param)\n",
        "clf.fit(X_train,y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/model_selection/_validation.py:372: FitFailedWarning: \n",
            "11 fits failed out of a total of 135.\n",
            "The score on these train-test partitions for these parameters will be set to nan.\n",
            "If these failures are not expected, you can try to debug them by setting error_score='raise'.\n",
            "\n",
            "Below are more details about the failures:\n",
            "--------------------------------------------------------------------------------\n",
            "11 fits failed with the following error:\n",
            "Traceback (most recent call last):\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/model_selection/_validation.py\", line 681, in _fit_and_score\n",
            "    estimator.fit(X_train, y_train, **fit_params)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/ensemble/_bagging.py\", line 269, in fit\n",
            "    return self._fit(X, y, self.max_samples, sample_weight=sample_weight)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/ensemble/_bagging.py\", line 407, in _fit\n",
            "    for i in range(n_jobs)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/parallel.py\", line 1043, in __call__\n",
            "    if self.dispatch_one_batch(iterator):\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/parallel.py\", line 861, in dispatch_one_batch\n",
            "    self._dispatch(tasks)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/parallel.py\", line 779, in _dispatch\n",
            "    job = self._backend.apply_async(batch, callback=cb)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/_parallel_backends.py\", line 208, in apply_async\n",
            "    result = ImmediateResult(func)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/_parallel_backends.py\", line 572, in __init__\n",
            "    self.results = batch()\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/parallel.py\", line 263, in __call__\n",
            "    for func, args, kwargs in self.items]\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/joblib/parallel.py\", line 263, in <listcomp>\n",
            "    for func, args, kwargs in self.items]\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/utils/fixes.py\", line 211, in __call__\n",
            "    return self.function(*args, **kwargs)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/ensemble/_bagging.py\", line 123, in _parallel_build_estimators\n",
            "    estimator.fit(X[:, features], y, sample_weight=curr_sample_weight)\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/naive_bayes.py\", line 246, in fit\n",
            "    X, y, np.unique(y), _refit=True, sample_weight=sample_weight\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/naive_bayes.py\", line 469, in _partial_fit\n",
            "    self.class_count_[i], self.theta_[i, :], self.var_[i, :], X_i, sw_i\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/sklearn/naive_bayes.py\", line 299, in _update_mean_variance\n",
            "    new_mu = np.average(X, axis=0, weights=sample_weight)\n",
            "  File \"<__array_function__ internals>\", line 6, in average\n",
            "  File \"/usr/local/lib/python3.7/dist-packages/numpy/lib/function_base.py\", line 410, in average\n",
            "    \"Weights sum to zero, can't be normalized\")\n",
            "ZeroDivisionError: Weights sum to zero, can't be normalized\n",
            "\n",
            "  warnings.warn(some_fits_failed_message, FitFailedWarning)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/model_selection/_search.py:972: UserWarning: One or more of the test scores are non-finite: [0.77929607 0.77184265 0.77060041 0.8426501  0.836853   0.83519669\n",
            " 0.79378882 0.78136646 0.77639752 0.84968944 0.85010352 0.85175983\n",
            " 0.85424431 0.85465839 0.85465839 0.85507246 0.85217391 0.85217391\n",
            " 0.27204969 0.28571429 0.28695652        nan        nan        nan\n",
            " 0.23478261 0.25341615 0.25755694]\n",
            "  category=UserWarning,\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(estimator=BaggingClassifier(),\n",
              "             param_grid={'base_estimator': [None,\n",
              "                                            LogisticRegression(max_iter=1000),\n",
              "                                            GaussianNB()],\n",
              "                         'max_samples': [1.0, 100, 0.5],\n",
              "                         'n_estimators': [5, 10, 15], 'random_state': [1]})"
            ]
          },
          "metadata": {},
          "execution_count": 62
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ar08uIH7Ucc7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ab72d713-8b6b-48df-f5d9-5531316af44f"
      },
      "source": [
        "best_model = clf.best_estimator_\n",
        "print('Parameters of Best Bagging Classifier Model: ')\n",
        "for k,i in clf.best_params_.items():\n",
        "  print('{}: {}'.format(k,i))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameters of Best Bagging Classifier Model: \n",
            "base_estimator: LogisticRegression(max_iter=1000)\n",
            "max_samples: 0.5\n",
            "n_estimators: 5\n",
            "random_state: 1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cUhEDsg4jCpL"
      },
      "source": [
        "train_pred,test_pred = pred(best_model,X_train,X_test) \n",
        "\n",
        "y_train_pred_af.append(train_pred)\n",
        "y_test_pred_af.append(test_pred)\n",
        "model_af.append(best_model)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xVXui_prYgNE"
      },
      "source": [
        "#### Model Comparison"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uj8yGWwE0Sqg",
        "outputId": "b6d39cff-9d59-4757-a386-567c01e9bbb5"
      },
      "source": [
        "model_bef\n",
        "# all models before tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[LogisticRegression(max_iter=1000),\n",
              " GaussianNB(),\n",
              " DecisionTreeClassifier(),\n",
              " RandomForestClassifier(),\n",
              " SVC(class_weight='balanced'),\n",
              " GradientBoostingClassifier(),\n",
              " AdaBoostClassifier(),\n",
              " BaggingClassifier()]"
            ]
          },
          "metadata": {},
          "execution_count": 65
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "smsYY7ty4XJB",
        "outputId": "071628d3-ab47-4904-b812-86c54765fe29"
      },
      "source": [
        "model_af\n",
        "# all models after tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[LogisticRegression(max_iter=1000),\n",
              " GaussianNB(),\n",
              " DecisionTreeClassifier(max_features='auto', max_leaf_nodes=2, random_state=1),\n",
              " RandomForestClassifier(min_impurity_decrease=0, min_samples_leaf=0.01,\n",
              "                        min_samples_split=0.05, n_estimators=10),\n",
              " SVC(C=2, class_weight='balanced', degree=5, gamma='auto', kernel='poly'),\n",
              " GradientBoostingClassifier(max_depth=1, min_samples_leaf=0.03, random_state=1),\n",
              " AdaBoostClassifier(base_estimator=LogisticRegression(), n_estimators=10,\n",
              "                    random_state=1),\n",
              " BaggingClassifier(base_estimator=LogisticRegression(max_iter=1000),\n",
              "                   max_samples=0.5, n_estimators=5, random_state=1)]"
            ]
          },
          "metadata": {},
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yUAOVm2Fa5xX"
      },
      "source": [
        "# Apply classification metrics such as confusion matrix, precision, recall, f-measure etc. \n",
        "def classification_metrics(y_true,y_pred,beta=1,ret=[]):\n",
        "  import numpy as np\n",
        "  from sklearn.metrics import classification_report, confusion_matrix, precision_recall_fscore_support, fbeta_score, accuracy_score, roc_auc_score\n",
        "  np.seterr(divide='ignore', invalid='ignore')\n",
        "\n",
        "  classification_report = classification_report(y_true,y_pred)\n",
        "  confusion_matrix = confusion_matrix(y_true,y_pred)\n",
        "\n",
        "  FP = confusion_matrix.sum(axis=0) - np.diag(confusion_matrix)  \n",
        "  FN = confusion_matrix.sum(axis=1) - np.diag(confusion_matrix)\n",
        "  TP = np.diag(confusion_matrix)\n",
        "  TN = confusion_matrix.sum() - (FP + FN + TP)\n",
        "  FPR = FP/(FP+TN) # aka Type 1 Error, Type I Error\n",
        "  FNR = FN/(FN+TP) # aka Type 2 Error, Type II Error \n",
        "  TPR = TP/(TP+FN) # aka Recall, Sensitivity\n",
        "  TNR = TN/(TN+FP) # aka Specificity\n",
        "\n",
        "  precision, TPR, f1_score, support = precision_recall_fscore_support(y_true,y_pred,zero_division=0)\n",
        "  try: accuracy = accuracy_score(y_true,y_pred)\n",
        "  except: accuracy = 0\n",
        "  try: roc_auc_score = roc_auc_score(y_true,y_pred)\n",
        "  except: roc_auc_score = 0\n",
        "\n",
        "  if ret == 'all': \n",
        "    beta = 2 \n",
        "    ret = ['classification_report','confusion_matrix','fpr','fnr','tpr','tnr','precision','f1_score','fbeta_score','accuracy','auc']\n",
        "  \n",
        "  scores=[]\n",
        "  for i in ret:\n",
        "    if i == 'classification_report': print('Classification Report:\\n',classification_report);\n",
        "    elif i == 'confusion_matrix': scores.append(confusion_matrix);\n",
        "    elif i in ('fpr','type_1_error','type_I_error'): scores.append(FPR);\n",
        "    elif i in ('fnr','type_2_error','type_II_error'): scores.append(FNR);\n",
        "    elif i in ('tpr','recall','sensitivity'): scores.append(TPR);\n",
        "    elif i in ('tnr','specificity'): scores.append(TNR)\n",
        "    elif i == 'precision': scores.append(precision)\n",
        "    elif i == 'f1_score': scores.append(f1_score)\n",
        "    elif i == 'fbeta_score': \n",
        "      try: scores.append(fbeta_score(y_true,y_pred,beta=beta,zero_division=1))\n",
        "      except: scores.append(0)\n",
        "    elif i in ('accuracy','accuracy_score'): scores.append(accuracy)\n",
        "    elif i in ('roc_auc','roc_auc_score','auc','auc_score'): scores.append(roc_auc_score)\n",
        "  if not scores: return None\n",
        "  return scores\n",
        "\n",
        "#def_metrics = classification_metrics(y_test,y_pred,ret='all')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WLKVN1wUktU1"
      },
      "source": [
        "##### training accuracy scores before hyperparameter tuning"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lUMNxvkbA0sK",
        "outputId": "4b3c9a76-0917-4fc6-95d2-07d35d86913b"
      },
      "source": [
        "toot_train=[]\n",
        "for y_pred in y_test_pred_bef:\n",
        "  toot_train.append(classification_metrics(y_test,y_pred,ret=['accuracy_score']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9rfrjGuOA0sL",
        "outputId": "30d00fdf-8a7a-46ac-fca0-c6bfac8cd519"
      },
      "source": [
        "toot_train\n",
        "# train accuracy before tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[0.8449131513647643],\n",
              " [0.6240694789081885],\n",
              " [0.7382133995037221],\n",
              " [0.749379652605459],\n",
              " [0.5620347394540943],\n",
              " [0.7568238213399504],\n",
              " [0.8449131513647643],\n",
              " [0.750620347394541]]"
            ]
          },
          "metadata": {},
          "execution_count": 69
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "65ZzwCHmkqij"
      },
      "source": [
        "##### testing accuracy scores before hyperparameter tuning"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "znuhy2TH3q01",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "dad3a5c5-ba09-44d8-ea0f-2078792adbe1"
      },
      "source": [
        "toot=[]\n",
        "for y_pred in y_test_pred_bef:\n",
        "  toot.append(classification_metrics(y_test,y_pred,ret=['accuracy_score']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TcUmYZtx33wV",
        "outputId": "5d41ea0c-99f7-4ba6-8d5c-918d0f46a07b"
      },
      "source": [
        "toot\n",
        "# accuracy before tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[0.8449131513647643],\n",
              " [0.6240694789081885],\n",
              " [0.7382133995037221],\n",
              " [0.749379652605459],\n",
              " [0.5620347394540943],\n",
              " [0.7568238213399504],\n",
              " [0.8449131513647643],\n",
              " [0.750620347394541]]"
            ]
          },
          "metadata": {},
          "execution_count": 71
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mJGtPB1HkmFv"
      },
      "source": [
        "##### training accuracy scores after hyperparameter tuning"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Sq1skE-tAk3N",
        "outputId": "06b99163-38d2-4256-861d-2e593763b2fd"
      },
      "source": [
        "something_train=[]\n",
        "for y_pred in y_test_pred_af:\n",
        "  something_train.append(classification_metrics(y_test,y_pred,ret=['accuracy_score']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TWwdjqYTAubE",
        "outputId": "53476547-d1f9-48a7-fb9d-42b10e2c10df"
      },
      "source": [
        "something_train\n",
        "# train scores after tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[0.8449131513647643],\n",
              " [0.6240694789081885],\n",
              " [0.8535980148883374],\n",
              " [0.8535980148883374],\n",
              " [0.7816377171215881],\n",
              " [0.8349875930521092],\n",
              " [0.8498759305210918],\n",
              " [0.8548387096774194]]"
            ]
          },
          "metadata": {},
          "execution_count": 73
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FORs2Lsgkfa9"
      },
      "source": [
        "##### testing accuracy scores after hyperparameter tuning"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oXuuJI-213xb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0fe47af6-1bec-43c9-a991-221eabd0c896"
      },
      "source": [
        "something=[]\n",
        "for y_pred in y_test_pred_af:\n",
        "  something.append(classification_metrics(y_test,y_pred,ret=['accuracy_score']))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3O9Hca_F29Dx",
        "outputId": "a0ab3c1b-bb82-40e4-f927-3de3388cc7bc"
      },
      "source": [
        "something\n",
        "# test scores after tuning"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[[0.8449131513647643],\n",
              " [0.6240694789081885],\n",
              " [0.8535980148883374],\n",
              " [0.8535980148883374],\n",
              " [0.7816377171215881],\n",
              " [0.8349875930521092],\n",
              " [0.8498759305210918],\n",
              " [0.8548387096774194]]"
            ]
          },
          "metadata": {},
          "execution_count": 75
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jNr85A7ZkO6D"
      },
      "source": [
        "##### testing f1-scores before hyperparameter tuning "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ynjQTLm8JwWo",
        "outputId": "deff9149-f9a7-4d2b-84b6-35343abc81e5"
      },
      "source": [
        "f1_test_bef = []\n",
        "for i in range(len(y_test_pred_bef)): \n",
        "  f1_test_bef.append(classification_metrics(y_test,y_test_pred_bef[i],ret=['f1_score'])[0])\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "abQPWCiEt85K",
        "outputId": "fe97f6f0-0166-41ff-ad9a-26e21f5fa55a"
      },
      "source": [
        "f1_test_bef"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[array([0.24      , 0.36363636, 0.91862069, 0.13333333]),\n",
              " array([0.34666667, 0.36363636, 0.73480176, 0.2310231 ]),\n",
              " array([0.18518519, 0.16      , 0.85059172, 0.02469136]),\n",
              " array([0.05      , 0.03846154, 0.85631769, 0.        ]),\n",
              " array([0.26315789, 0.32432432, 0.6493756 , 0.23529412]),\n",
              " array([0.13043478, 0.07407407, 0.86251809, 0.02631579]),\n",
              " array([0.4       , 0.36111111, 0.91947694, 0.        ]),\n",
              " array([0.12765957, 0.09009009, 0.85756026, 0.07058824])]"
            ]
          },
          "metadata": {},
          "execution_count": 142
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-HnoVVY7kHcz"
      },
      "source": [
        "##### testing f1-scores after hyperparameter tuning "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zisq0lFjQ82N",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d1efdec8-972b-4930-89ec-3d4ac56d38e8"
      },
      "source": [
        "f1_test_af = []\n",
        "for i in range(len(y_test_pred_af)): \n",
        "  f1_test_af.append(classification_metrics(y_test,y_test_pred_af[i],ret=['f1_score'])[0])\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_ranking.py:546: FutureWarning: Arrays of bytes/strings is being converted to decimal numbers if dtype='numeric'. This behavior is deprecated in 0.24 and will be removed in 1.1 (renaming of 0.26). Please convert your data to numeric values explicitly instead.\n",
            "  y_score = check_array(y_score, ensure_2d=False)\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1308: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DQsDiApCRWki",
        "outputId": "b75b2b9d-2055-46ed-b858-1141438396f8"
      },
      "source": [
        "f1_test_af"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[array([0.24      , 0.36363636, 0.91862069, 0.13333333]),\n",
              " array([0.34666667, 0.36363636, 0.73480176, 0.2310231 ]),\n",
              " array([0.        , 0.        , 0.92461641, 0.        ]),\n",
              " array([0.        , 0.30136986, 0.91983696, 0.        ]),\n",
              " array([0.32835821, 0.42767296, 0.86124402, 0.40909091]),\n",
              " array([0.15384615, 0.2745098 , 0.90692792, 0.03636364]),\n",
              " array([0.        , 0.        , 0.92317969, 0.        ]),\n",
              " array([0.        , 0.        , 0.92317969, 0.        ])]"
            ]
          },
          "metadata": {},
          "execution_count": 144
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H966B0FIkbUN"
      },
      "source": [
        "##### stratified k-fold for k=10\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_N895iWFmeEt",
        "outputId": "0d174cda-3114-47df-90a9-5ea7f7b6b168"
      },
      "source": [
        "from sklearn.model_selection import StratifiedKFold\n",
        "\n",
        "kX = np.array(X)\n",
        "ky = np.array(y)\n",
        "\n",
        "fin_acc=[]\n",
        "\n",
        "for lr in model_af:\n",
        "  skf = StratifiedKFold(n_splits=10, shuffle=True, random_state=1)\n",
        "  lst_accu_stratified = []\n",
        "  for train_index, test_index in skf.split(kX, ky):\n",
        "      x_train_fold, x_test_fold = kX[train_index], kX[test_index]\n",
        "      y_train_fold, y_test_fold = ky[train_index], ky[test_index]\n",
        "      lr.fit(x_train_fold, y_train_fold)\n",
        "      lst_accu_stratified.append(lr.score(x_test_fold, y_test_fold))\n",
        "  fin_acc.append(np.mean(lst_accu_stratified))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/linear_model/_logistic.py:818: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG,\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DsEuq1IwoDPz",
        "outputId": "cadd7e16-92fb-41a8-aaee-b3762c4acd94"
      },
      "source": [
        "fin_acc\n",
        "# max = 0.8547035747937619"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.8516037536295983,\n",
              " 0.5631675095667559,\n",
              " 0.8547035747937619,\n",
              " 0.8543930157875506,\n",
              " 0.7655933311539719,\n",
              " 0.8404245908889874,\n",
              " 0.8537718977751283,\n",
              " 0.8531527027286888]"
            ]
          },
          "metadata": {},
          "execution_count": 148
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uFTbwcldkSmi"
      },
      "source": [
        "##### Best Model "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "l5IGmEs43D1i",
        "outputId": "ae820680-5298-4b93-b917-150a922b5f24"
      },
      "source": [
        "model_af[something.index(max(something))]\n",
        "# model with best accuracy "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "DecisionTreeClassifier(max_features='auto', max_leaf_nodes=2, random_state=1)"
            ]
          },
          "metadata": {},
          "execution_count": 139
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GRn0srhLFyYT",
        "outputId": "fb94ac84-483d-4f09-c8f0-37028dfb3abd"
      },
      "source": [
        "max(something)[0]*100\n",
        "# best accuracy"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "85.98014888337468"
            ]
          },
          "metadata": {},
          "execution_count": 140
        }
      ]
    }
  ]
}