Metrica_personalizada_para_ML.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Metrica personalizada para ML",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPGuwu+Bmxr9V/8Qu7JFCyT",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ricardorocha86/Python-para-Machine-Learning/blob/master/Metrica_personalizada_para_ML.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7utdAe5d2sLN"
      },
      "source": [
        "![texto alternativo](https://github.com/gitflai/imagens/blob/main/capa.jpg?raw=true)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VCPmNK__4ryO"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Z4Go_M720Y4"
      },
      "source": [
        "## **Métricas para Classificação Binária**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wwhT9yFV4vuS"
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xwzWV5roWp-R"
      },
      "source": [
        "url1 = 'https://raw.githubusercontent.com/ricardorocha86/Datasets/master/Titanic/train.csv' \n",
        "titanic = pd.read_csv(url1) "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eUeh-zd34wLk"
      },
      "source": [
        "teste1 = titanic['Sex'].map({'male': 0, 'female': 1})"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A6xlGADLpmKd"
      },
      "source": [
        "from sklearn.metrics import confusion_matrix "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lH-Zb7bYvgV8"
      },
      "source": [
        "mc = confusion_matrix(titanic['Survived'], teste1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "khCtOQ6EyFKG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "398a7174-d45c-40da-c32d-0f26ce2b270e"
      },
      "source": [
        "mc"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[468,  81],\n",
              "       [109, 233]])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 104
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tC8k2TWXtaCu",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a831bf42-8f69-4522-a9d1-1220a861ad24"
      },
      "source": [
        "vn, fp, fn, vp = mc.flatten()\n",
        "\n",
        "print('VP: {}'.format(vp))\n",
        "print('VN: {}'.format(vn))\n",
        "print('FP: {}'.format(fp))\n",
        "print('FN: {}'.format(fn))\n",
        "print('\\nTOTAL: {}'.format(fn+fp+vn+vp))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "VP: 233\n",
            "VN: 468\n",
            "FP: 81\n",
            "FN: 109\n",
            "\n",
            "TOTAL: 891\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j7qIlvRhxi54"
      },
      "source": [
        "def MetricasClassBinaria(observado, predito, custo = None):\n",
        "    mc = confusion_matrix(observado, predito)\n",
        "    vn, fp, fn, vp = mc.flatten()\n",
        "\n",
        "    pre = observado.mean()\n",
        "    acc = (vn + vp)/(vn + fp + fn + vp) \n",
        "    sen = vp/(vp + fn)\n",
        "    esp = vn/(vn + fp)\n",
        "    vpp = vp/(vp + fp)\n",
        "    vpn = vn/(vn + fn) \n",
        "    mcc = (vp*vn - fp*fn)/np.sqrt((vp + fp)*(vp + fn)*(vn + fp)*(vn + fn))\n",
        "    f1s = 2*vpp*sen/(vpp + sen) \n",
        "    efi = (sen + esp)/2\n",
        "\n",
        "    print('  Prevalência de Positivos: {:.4f}'.format(pre))\n",
        "    print('                  Acurácia: {:.4f}'.format(acc))\n",
        "    print('             Sensibilidade: {:.4f}'.format(sen))\n",
        "    print('            Especificidade: {:.4f}'.format(esp))\n",
        "    print('  Valor Preditivo Positivo: {:.4f}'.format(vpp))\n",
        "    print('  Valor Preditivo Negativo: {:.4f}'.format(vpn))\n",
        "    print('Coef. de Corr. de Matthews: {:.4f}'.format(mcc))\n",
        "    print('                  F1-Score: {:.4f}'.format(f1s))\n",
        "    print('                Eficiência: {:.4f}'.format(efi))\n",
        " \n",
        "\n",
        "    metricas = {'prevalência': pre, 'acurácia': acc, 'sensibilidade': sen, \n",
        "            'especificidade': esp, 'vpp': vpp, 'vpn': vpn, 'mcc': mcc,\n",
        "            'f1-score': f1s, 'eficiência': efi}\n",
        "\n",
        "    if custo is not None:\n",
        "        cost = (mc * np.array(custo).reshape(2,2)).sum()\n",
        "        print('\\n                     Custo: {}'.format(cost))\n",
        "        metricas['custo'] = cost\n",
        "\n",
        "    return metricas"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "f1jm1adY1PLI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "49539ce9-9a53-4181-dafb-1bd0602f49f5"
      },
      "source": [
        "metricas = MetricasClassBinaria(titanic['Survived'], teste1, [1, -1, 20, 2])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.3838\n",
            "                  Acurácia: 0.7868\n",
            "             Sensibilidade: 0.6813\n",
            "            Especificidade: 0.8525\n",
            "  Valor Preditivo Positivo: 0.7420\n",
            "  Valor Preditivo Negativo: 0.8111\n",
            "Coef. de Corr. de Matthews: 0.5434\n",
            "                  F1-Score: 0.7104\n",
            "                Eficiência: 0.7669\n",
            "\n",
            "                     Custo: 3033\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Iw1pM8ho4CQw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b28e9d87-cbad-4052-c221-312ea18a60b1"
      },
      "source": [
        "metricas"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'acurácia': 0.7867564534231201,\n",
              " 'custo': 3033,\n",
              " 'eficiência': 0.7668727830505224,\n",
              " 'especificidade': 0.8524590163934426,\n",
              " 'f1-score': 0.7103658536585367,\n",
              " 'mcc': 0.5433513806577551,\n",
              " 'prevalência': 0.3838383838383838,\n",
              " 'sensibilidade': 0.6812865497076024,\n",
              " 'vpn': 0.8110918544194108,\n",
              " 'vpp': 0.7420382165605095}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 108
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WrFVEL8E6BCm"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Kt3NH9w2QbrO"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zmJAvGN0QceR"
      },
      "source": [
        "## **Exemplo Final: Modelagem de Churn**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Az78uyadQhjI",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 226
        },
        "outputId": "7b310365-7d9a-47c4-ea33-15add2bcc5b8"
      },
      "source": [
        "url = 'https://raw.githubusercontent.com/ricardorocha86/Datasets/master/Churn/churn.csv'\n",
        "churn = pd.read_csv(url)\n",
        "churn.head()"
      ],
      "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>RowNumber</th>\n",
              "      <th>CustomerId</th>\n",
              "      <th>Surname</th>\n",
              "      <th>CreditScore</th>\n",
              "      <th>Geography</th>\n",
              "      <th>Gender</th>\n",
              "      <th>Age</th>\n",
              "      <th>Tenure</th>\n",
              "      <th>Balance</th>\n",
              "      <th>NumOfProducts</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>EstimatedSalary</th>\n",
              "      <th>Exited</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>15634602</td>\n",
              "      <td>Hargrave</td>\n",
              "      <td>619</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>2</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>101348.88</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2</td>\n",
              "      <td>15647311</td>\n",
              "      <td>Hill</td>\n",
              "      <td>608</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>41</td>\n",
              "      <td>1</td>\n",
              "      <td>83807.86</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>112542.58</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>15619304</td>\n",
              "      <td>Onio</td>\n",
              "      <td>502</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>8</td>\n",
              "      <td>159660.80</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>113931.57</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>15701354</td>\n",
              "      <td>Boni</td>\n",
              "      <td>699</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>39</td>\n",
              "      <td>1</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>93826.63</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>5</td>\n",
              "      <td>15737888</td>\n",
              "      <td>Mitchell</td>\n",
              "      <td>850</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>43</td>\n",
              "      <td>2</td>\n",
              "      <td>125510.82</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>79084.10</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   RowNumber  CustomerId   Surname  ...  IsActiveMember EstimatedSalary Exited\n",
              "0          1    15634602  Hargrave  ...               1       101348.88      1\n",
              "1          2    15647311      Hill  ...               1       112542.58      0\n",
              "2          3    15619304      Onio  ...               0       113931.57      1\n",
              "3          4    15701354      Boni  ...               0        93826.63      0\n",
              "4          5    15737888  Mitchell  ...               1        79084.10      0\n",
              "\n",
              "[5 rows x 14 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 110
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5SfHWbEzRZk6"
      },
      "source": [
        "### Análise Exploratória de Dados Básica"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w9x-i9ofRgoJ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5a16db14-b5d7-427c-a69e-2c9168672f38"
      },
      "source": [
        "churn.info()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 10000 entries, 0 to 9999\n",
            "Data columns (total 14 columns):\n",
            " #   Column           Non-Null Count  Dtype  \n",
            "---  ------           --------------  -----  \n",
            " 0   RowNumber        10000 non-null  int64  \n",
            " 1   CustomerId       10000 non-null  int64  \n",
            " 2   Surname          10000 non-null  object \n",
            " 3   CreditScore      10000 non-null  int64  \n",
            " 4   Geography        10000 non-null  object \n",
            " 5   Gender           10000 non-null  object \n",
            " 6   Age              10000 non-null  int64  \n",
            " 7   Tenure           10000 non-null  int64  \n",
            " 8   Balance          10000 non-null  float64\n",
            " 9   NumOfProducts    10000 non-null  int64  \n",
            " 10  HasCrCard        10000 non-null  int64  \n",
            " 11  IsActiveMember   10000 non-null  int64  \n",
            " 12  EstimatedSalary  10000 non-null  float64\n",
            " 13  Exited           10000 non-null  int64  \n",
            "dtypes: float64(2), int64(9), object(3)\n",
            "memory usage: 1.1+ MB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z9CL4Tj7RlqB",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "03f5aca8-1a77-4bae-b1fd-3a6ad1dcf7e8"
      },
      "source": [
        "churn.isnull().sum()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "RowNumber          0\n",
              "CustomerId         0\n",
              "Surname            0\n",
              "CreditScore        0\n",
              "Geography          0\n",
              "Gender             0\n",
              "Age                0\n",
              "Tenure             0\n",
              "Balance            0\n",
              "NumOfProducts      0\n",
              "HasCrCard          0\n",
              "IsActiveMember     0\n",
              "EstimatedSalary    0\n",
              "Exited             0\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 112
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LyhZJNFBHPAf"
      },
      "source": [
        "!pip install sweetviz"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t_BLjf0eHOCm"
      },
      "source": [
        "import sweetviz "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XHzabtjKJzBg"
      },
      "source": [
        "eda = sweetviz.compare_intra(churn, churn['Exited'] == 0, ['Cliente', 'Ex-Cliente'])\n",
        "eda.show_html()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yX9QZY5iRowi"
      },
      "source": [
        "### Pré-processamento dos Dados"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xPOK_S9mSfr6"
      },
      "source": [
        "#### 1. Novas Variáveis"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Cmyw0RxfUm4g"
      },
      "source": [
        "def aux1(x):\n",
        "    if x == 0:\n",
        "        return 'zero'\n",
        "    elif x < 100000:\n",
        "        return '0-100k'\n",
        "    else:\n",
        "        return '100k+'\n",
        "\n",
        "churn['Saldo'] = churn['Balance'].apply(aux1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7EHU-YyMVZP-"
      },
      "source": [
        "def aux2(x):\n",
        "    if x < 3:\n",
        "        return '0-2'\n",
        "    elif x < 8:\n",
        "        return '3-7'\n",
        "    else:\n",
        "        return '8+'\n",
        "\n",
        "churn['Tempo'] = churn['Tenure'].apply(aux2)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O8dgtPQZXBiA"
      },
      "source": [
        "def aux3(x):\n",
        "    if x == 1:\n",
        "        return '1'\n",
        "    elif x == 2:\n",
        "        return '2'\n",
        "    else:\n",
        "        return '3+'\n",
        "\n",
        "churn['Produtos'] = churn['NumOfProducts'].apply(aux3)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TK4cof6XWjpW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 226
        },
        "outputId": "be06a8e0-4948-4fab-967d-d84a4836fe7d"
      },
      "source": [
        "churn.head()"
      ],
      "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>RowNumber</th>\n",
              "      <th>CustomerId</th>\n",
              "      <th>Surname</th>\n",
              "      <th>CreditScore</th>\n",
              "      <th>Geography</th>\n",
              "      <th>Gender</th>\n",
              "      <th>Age</th>\n",
              "      <th>Tenure</th>\n",
              "      <th>Balance</th>\n",
              "      <th>NumOfProducts</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>EstimatedSalary</th>\n",
              "      <th>Exited</th>\n",
              "      <th>Saldo</th>\n",
              "      <th>Tempo</th>\n",
              "      <th>Produtos</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>15634602</td>\n",
              "      <td>Hargrave</td>\n",
              "      <td>619</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>2</td>\n",
              "      <td>0.00</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>101348.88</td>\n",
              "      <td>1</td>\n",
              "      <td>zero</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2</td>\n",
              "      <td>15647311</td>\n",
              "      <td>Hill</td>\n",
              "      <td>608</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>41</td>\n",
              "      <td>1</td>\n",
              "      <td>83807.86</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>112542.58</td>\n",
              "      <td>0</td>\n",
              "      <td>0-100k</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>15619304</td>\n",
              "      <td>Onio</td>\n",
              "      <td>502</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>42</td>\n",
              "      <td>8</td>\n",
              "      <td>159660.80</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>113931.57</td>\n",
              "      <td>1</td>\n",
              "      <td>100k+</td>\n",
              "      <td>8+</td>\n",
              "      <td>3+</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>15701354</td>\n",
              "      <td>Boni</td>\n",
              "      <td>699</td>\n",
              "      <td>France</td>\n",
              "      <td>Female</td>\n",
              "      <td>39</td>\n",
              "      <td>1</td>\n",
              "      <td>0.00</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>93826.63</td>\n",
              "      <td>0</td>\n",
              "      <td>zero</td>\n",
              "      <td>0-2</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>5</td>\n",
              "      <td>15737888</td>\n",
              "      <td>Mitchell</td>\n",
              "      <td>850</td>\n",
              "      <td>Spain</td>\n",
              "      <td>Female</td>\n",
              "      <td>43</td>\n",
              "      <td>2</td>\n",
              "      <td>125510.82</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>79084.10</td>\n",
              "      <td>0</td>\n",
              "      <td>100k+</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   RowNumber  CustomerId   Surname  CreditScore  ... Exited   Saldo  Tempo  Produtos\n",
              "0          1    15634602  Hargrave          619  ...      1    zero    0-2         1\n",
              "1          2    15647311      Hill          608  ...      0  0-100k    0-2         1\n",
              "2          3    15619304      Onio          502  ...      1   100k+     8+        3+\n",
              "3          4    15701354      Boni          699  ...      0    zero    0-2         2\n",
              "4          5    15737888  Mitchell          850  ...      0   100k+    0-2         1\n",
              "\n",
              "[5 rows x 17 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 119
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZrkKYUgRSZ6g"
      },
      "source": [
        "#### 2. Dropando Variáveis"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-zU0VcXWSe1p"
      },
      "source": [
        "churn.drop(['RowNumber', 'CustomerId', 'Surname', 'Balance', 'Tenure', 'NumOfProducts'], axis = 1, inplace = True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gqw6pDvjSd8J"
      },
      "source": [
        "churn.head()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ph-cI3VQSfr5"
      },
      "source": [
        "#### 3. Dummificações"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EYJ830t8ShXE"
      },
      "source": [
        "churn = pd.get_dummies(churn, columns = ['Geography', 'Gender', 'Saldo', 'Tempo', 'Produtos'], drop_first = True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xt1ZloEJl9xW"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EDUODYnyShZ4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 226
        },
        "outputId": "6e634934-53e4-41c0-bb62-5ed9d29b4a42"
      },
      "source": [
        "churn.head()"
      ],
      "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>CreditScore</th>\n",
              "      <th>Age</th>\n",
              "      <th>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>EstimatedSalary</th>\n",
              "      <th>Exited</th>\n",
              "      <th>Geography_Germany</th>\n",
              "      <th>Geography_Spain</th>\n",
              "      <th>Gender_Male</th>\n",
              "      <th>Saldo_100k+</th>\n",
              "      <th>Saldo_zero</th>\n",
              "      <th>Tempo_3-7</th>\n",
              "      <th>Tempo_8+</th>\n",
              "      <th>Produtos_2</th>\n",
              "      <th>Produtos_3+</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>619</td>\n",
              "      <td>42</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>101348.88</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>608</td>\n",
              "      <td>41</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>112542.58</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>502</td>\n",
              "      <td>42</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>113931.57</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>699</td>\n",
              "      <td>39</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>93826.63</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>850</td>\n",
              "      <td>43</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>79084.10</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   CreditScore  Age  HasCrCard  ...  Tempo_8+  Produtos_2  Produtos_3+\n",
              "0          619   42          1  ...         0           0            0\n",
              "1          608   41          0  ...         0           0            0\n",
              "2          502   42          1  ...         1           0            1\n",
              "3          699   39          0  ...         0           1            0\n",
              "4          850   43          1  ...         0           0            0\n",
              "\n",
              "[5 rows x 15 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 122
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vzM911q_Sf-Q"
      },
      "source": [
        "#### 4. Padronizações"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DYTMt8Pa5FVr"
      },
      "source": [
        "from sklearn.preprocessing import StandardScaler\n",
        "sc = StandardScaler()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Nl9NAL9c77_k"
      },
      "source": [
        "aux0 = sc.fit_transform(churn[['CreditScore', 'Age', 'EstimatedSalary']])\n",
        "aux1 = pd.DataFrame(aux0, columns = ['Escore', 'Idade', 'Salario'])\n",
        "churn = pd.concat([churn, aux1], axis = 1)\n",
        "churn.drop(['CreditScore', 'Age', 'EstimatedSalary'], axis = 1, inplace = True) "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EHWIRZ6CX7tm",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 226
        },
        "outputId": "890a29d1-0878-4e60-bb4e-ad06c89ddff3"
      },
      "source": [
        "churn.head()"
      ],
      "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>HasCrCard</th>\n",
              "      <th>IsActiveMember</th>\n",
              "      <th>Exited</th>\n",
              "      <th>Geography_Germany</th>\n",
              "      <th>Geography_Spain</th>\n",
              "      <th>Gender_Male</th>\n",
              "      <th>Saldo_100k+</th>\n",
              "      <th>Saldo_zero</th>\n",
              "      <th>Tempo_3-7</th>\n",
              "      <th>Tempo_8+</th>\n",
              "      <th>Produtos_2</th>\n",
              "      <th>Produtos_3+</th>\n",
              "      <th>Escore</th>\n",
              "      <th>Idade</th>\n",
              "      <th>Salario</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>-0.326221</td>\n",
              "      <td>0.293517</td>\n",
              "      <td>0.021886</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>-0.440036</td>\n",
              "      <td>0.198164</td>\n",
              "      <td>0.216534</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>-1.536794</td>\n",
              "      <td>0.293517</td>\n",
              "      <td>0.240687</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0.501521</td>\n",
              "      <td>0.007457</td>\n",
              "      <td>-0.108918</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>2.063884</td>\n",
              "      <td>0.388871</td>\n",
              "      <td>-0.365276</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   HasCrCard  IsActiveMember  Exited  ...    Escore     Idade   Salario\n",
              "0          1               1       1  ... -0.326221  0.293517  0.021886\n",
              "1          0               1       0  ... -0.440036  0.198164  0.216534\n",
              "2          1               0       1  ... -1.536794  0.293517  0.240687\n",
              "3          0               0       0  ...  0.501521  0.007457 -0.108918\n",
              "4          1               1       0  ...  2.063884  0.388871 -0.365276\n",
              "\n",
              "[5 rows x 15 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 125
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UoAeQ6-eYGL0"
      },
      "source": [
        "## **Modelagem**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AkyFUMP1YMH2"
      },
      "source": [
        "Nesse caso, temos **dois objetivos**:\n",
        "\n",
        "- Selecionar um bom modelo que se ajuste a esses dados\n",
        "- Estimar o erro de generalização do modelo\n",
        "\n",
        "Para isso, vamos dividir os dados em **treinamento (90%)** e **validação (10%)**. Como temos 10000 observações no conjunto de dados, a divisão pode ser mais desbalanceada, de forma que dessa maneira ainda sobra 1000 elementos no conjunto de validação.\n",
        "\n",
        "Vamos utilizar **validação cruzada no conjunto de dados de treinamento** para encontrar os melhores parâmetros para os modelos propostos e, por fim, verificaremos o **erro de generalização dos modelos na amostra final de validação**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J5zaNTFWZZ-D"
      },
      "source": [
        "#Fazendo a Divisão em Teste e Treinamento\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "Xtreino, Xvalid, Ytreino, Yvalid = train_test_split(churn.drop('Exited', axis = 1), \n",
        "                                                    churn['Exited'], \n",
        "                                                    test_size = 1000)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "X3qBFvVIZyOB"
      },
      "source": [
        "# Configuração da Validação Cruzada\n",
        "from sklearn.model_selection import RepeatedStratifiedKFold\n",
        "from sklearn.model_selection import cross_val_score, cross_validate\n",
        "\n",
        "kf = RepeatedStratifiedKFold(n_splits = 10, n_repeats = 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EqIIvcgFZznJ"
      },
      "source": [
        "# Importando os Modelos\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.tree import DecisionTreeClassifier\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.ensemble import RandomForestClassifier"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "E1LlrINDexHB"
      },
      "source": [
        "# Definindo a métrica de interesse \n",
        "metrica = 'f1'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zyz9DnNvJO38"
      },
      "source": [
        "### Regressão Logística "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iNG_hF-rJSmu",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "95dcf95c-dc26-4cfd-f2a2-039e6e544d40"
      },
      "source": [
        "# Verificando a Performance da Regressão Logística\n",
        "\n",
        "reglog = cross_val_score(LogisticRegression(), Xtreino, Ytreino, cv = kf, scoring = metrica)\n",
        " \n",
        "print('F1-Score: {:.2f}%'.format(100*sens_reglog.mean()))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "F1-Score: 49.58%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Pq-NEx6_bwrr",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ba3a8c05-cd22-42f9-df55-f35c7856439b"
      },
      "source": [
        "reglog = LogisticRegression()\n",
        "reglog.fit(Xtreino, Ytreino)\n",
        "pred_reglog = reglog.predict(Xvalid)\n",
        "\n",
        "met_reglog = MetricasClassBinaria(Yvalid, pred_reglog)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8270\n",
            "             Sensibilidade: 0.3476\n",
            "            Especificidade: 0.9544\n",
            "  Valor Preditivo Positivo: 0.6697\n",
            "  Valor Preditivo Negativo: 0.8462\n",
            "Coef. de Corr. de Matthews: 0.3948\n",
            "                  F1-Score: 0.4577\n",
            "                Eficiência: 0.6510\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CHQoNHN1I-jM"
      },
      "source": [
        "### K Vizinhos Mais Próximos"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1dtlCS4AAt8p",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0b648e82-2645-44ae-df7f-60e3c3301484"
      },
      "source": [
        "# Encontrando o melhor valor de K através de Validação Cruzada K-FOLD \n",
        "\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "valores = np.arange(1, 51, 3)\n",
        "parametros = {'n_neighbors': valores}\n",
        "\n",
        "clf = GridSearchCV(KNeighborsClassifier(), param_grid = parametros, cv = kf, scoring = metrica)\n",
        "clf.fit(Xtreino, Ytreino)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(cv=RepeatedStratifiedKFold(n_repeats=1, n_splits=10, random_state=None),\n",
              "             error_score=nan,\n",
              "             estimator=KNeighborsClassifier(algorithm='auto', leaf_size=30,\n",
              "                                            metric='minkowski',\n",
              "                                            metric_params=None, n_jobs=None,\n",
              "                                            n_neighbors=5, p=2,\n",
              "                                            weights='uniform'),\n",
              "             iid='deprecated', n_jobs=None,\n",
              "             param_grid={'n_neighbors': array([ 1,  4,  7, 10, 13, 16, 19, 22, 25, 28, 31, 34, 37, 40, 43, 46, 49])},\n",
              "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
              "             scoring='f1', verbose=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 131
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jpx_mspcBTFU",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "20d00e01-2fcf-4bec-dfa0-423a8dd6c38b"
      },
      "source": [
        "# melhor escolha de K\n",
        "print(clf.cv_results_['mean_test_score'])\n",
        "print(clf.best_params_)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0.47581886 0.41771631 0.4783874  0.42991545 0.46580643 0.41750366\n",
            " 0.44888702 0.42070547 0.44379831 0.41752921 0.42884005 0.41247336\n",
            " 0.42863955 0.41068541 0.4199579  0.40106431 0.41025131]\n",
            "{'n_neighbors': 7}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o6NpAG4H8GDF",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ee1854eb-9153-41ee-90f4-907b0add8dd2"
      },
      "source": [
        "# verificando a performance do KNN para um dado valor de K\n",
        "\n",
        "knn5 = KNeighborsClassifier(n_neighbors = 7)\n",
        "knn5.fit(Xtreino, Ytreino)\n",
        "pred_knn5 = knn5.predict(Xvalid)\n",
        "\n",
        "met_knn5 = MetricasClassBinaria(Yvalid, pred_knn5)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8290\n",
            "             Sensibilidade: 0.3524\n",
            "            Especificidade: 0.9557\n",
            "  Valor Preditivo Positivo: 0.6789\n",
            "  Valor Preditivo Negativo: 0.8474\n",
            "Coef. de Corr. de Matthews: 0.4027\n",
            "                  F1-Score: 0.4639\n",
            "                Eficiência: 0.6540\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ob3dA8W9fuj5"
      },
      "source": [
        "### Árvore de Decisão"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IpR_G6CGJ1Rp",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8049ce22-ccca-4dea-d08e-a2b4a1970dab"
      },
      "source": [
        "# Encontrando o melhor valor do 'max_depth' através de Validação Cruzada K-FOLD\n",
        "\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "valores = np.arange(2, 15, 1)\n",
        "parametros = {'max_depth': valores}\n",
        "\n",
        "clf = GridSearchCV(DecisionTreeClassifier(), param_grid = parametros, cv = kf, scoring = metrica)\n",
        "clf.fit(Xtreino, Ytreino)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(cv=RepeatedStratifiedKFold(n_repeats=1, n_splits=10, random_state=None),\n",
              "             error_score=nan,\n",
              "             estimator=DecisionTreeClassifier(ccp_alpha=0.0, class_weight=None,\n",
              "                                              criterion='gini', max_depth=None,\n",
              "                                              max_features=None,\n",
              "                                              max_leaf_nodes=None,\n",
              "                                              min_impurity_decrease=0.0,\n",
              "                                              min_impurity_split=None,\n",
              "                                              min_samples_leaf=1,\n",
              "                                              min_samples_split=2,\n",
              "                                              min_weight_fraction_leaf=0.0,\n",
              "                                              presort='deprecated',\n",
              "                                              random_state=None,\n",
              "                                              splitter='best'),\n",
              "             iid='deprecated', n_jobs=None,\n",
              "             param_grid={'max_depth': array([ 2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14])},\n",
              "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
              "             scoring='f1', verbose=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 135
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r_6x0H0hf_XZ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "94cbd422-2b9c-4f72-fa17-49ed2837a28f"
      },
      "source": [
        "# melhor escolha de profundidade\n",
        "print(clf.cv_results_['mean_test_score'])\n",
        "print(clf.best_params_)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0.53241974 0.50570716 0.527137   0.56297677 0.53825227 0.55323726\n",
            " 0.54822276 0.54691536 0.53916868 0.52482955 0.5254693  0.5169665\n",
            " 0.5074141 ]\n",
            "{'max_depth': 5}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G3TFF5iyf_Xo",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1023d918-74ca-4586-9d17-cfb8587f177b"
      },
      "source": [
        "# verificando a performance do KNN para um dado valor de K\n",
        "\n",
        "arvore = DecisionTreeClassifier(max_depth = 5)\n",
        "arvore.fit(Xtreino, Ytreino)\n",
        "pred_arvore = arvore.predict(Xvalid)\n",
        "\n",
        "met_arvore = MetricasClassBinaria(Yvalid, pred_arvore)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8400\n",
            "             Sensibilidade: 0.3286\n",
            "            Especificidade: 0.9759\n",
            "  Valor Preditivo Positivo: 0.7841\n",
            "  Valor Preditivo Negativo: 0.8454\n",
            "Coef. de Corr. de Matthews: 0.4378\n",
            "                  F1-Score: 0.4631\n",
            "                Eficiência: 0.6523\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XXBFlpeMMsSv"
      },
      "source": [
        "### Random Forests"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QiV4ex62MsTA",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "93aef898-b8e9-4329-d271-18d2447c6dea"
      },
      "source": [
        "# Encontrando o melhor valor do 'max_depth' através de Validação Cruzada K-FOLD\n",
        "\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "valores = np.arange(2, 30, 3)\n",
        "parametros = {'max_depth': valores}\n",
        "\n",
        "clf = GridSearchCV(RandomForestClassifier(), param_grid = parametros, cv = kf, scoring = metrica)\n",
        "clf.fit(Xtreino, Ytreino)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(cv=RepeatedStratifiedKFold(n_repeats=1, n_splits=10, random_state=None),\n",
              "             error_score=nan,\n",
              "             estimator=RandomForestClassifier(bootstrap=True, ccp_alpha=0.0,\n",
              "                                              class_weight=None,\n",
              "                                              criterion='gini', max_depth=None,\n",
              "                                              max_features='auto',\n",
              "                                              max_leaf_nodes=None,\n",
              "                                              max_samples=None,\n",
              "                                              min_impurity_decrease=0.0,\n",
              "                                              min_impurity_split=None,\n",
              "                                              min_samples_leaf=1,\n",
              "                                              min_samples_split=2,\n",
              "                                              min_weight_fraction_leaf=0.0,\n",
              "                                              n_estimators=100, n_jobs=None,\n",
              "                                              oob_score=False,\n",
              "                                              random_state=None, verbose=0,\n",
              "                                              warm_start=False),\n",
              "             iid='deprecated', n_jobs=None,\n",
              "             param_grid={'max_depth': array([ 2,  5,  8, 11, 14, 17, 20, 23, 26, 29])},\n",
              "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
              "             scoring='f1', verbose=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 138
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ix5-Nlxng0WP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d55a7c28-099d-4058-ee01-71872b1dfcf2"
      },
      "source": [
        "# melhor escolha de profundidade\n",
        "print(clf.cv_results_['mean_test_score'])\n",
        "print(clf.best_params_)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0.10463274 0.43799048 0.5493966  0.56673273 0.57205899 0.57459513\n",
            " 0.56333482 0.56983306 0.57627725 0.5707966 ]\n",
            "{'max_depth': 26}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hzoK-jF9g3uT",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8856dae0-ef4c-4972-8491-cc83b4cd466f"
      },
      "source": [
        "# verificando a performance\n",
        "\n",
        "rf = RandomForestClassifier(max_depth = 26)\n",
        "rf.fit(Xtreino, Ytreino)\n",
        "pred_rf = rf.predict(Xvalid)\n",
        "\n",
        "met_rf = MetricasClassBinaria(Yvalid, pred_rf)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8460\n",
            "             Sensibilidade: 0.4238\n",
            "            Especificidade: 0.9582\n",
            "  Valor Preditivo Positivo: 0.7295\n",
            "  Valor Preditivo Negativo: 0.8622\n",
            "Coef. de Corr. de Matthews: 0.4754\n",
            "                  F1-Score: 0.5361\n",
            "                Eficiência: 0.6910\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1Np82CcjiOru"
      },
      "source": [
        "### Síntese dos Resultados"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3nG16DfQiQy4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "83dbdad5-8cf5-4b90-a450-a7354eaf81b1"
      },
      "source": [
        "# Regressão Logística\n",
        "reglog = LogisticRegression()\n",
        "reglog.fit(Xtreino, Ytreino)\n",
        "pred_reglog = reglog.predict(Xvalid)\n",
        "met_reglog = MetricasClassBinaria(Yvalid, pred_reglog, custo = [0, 1, 10, 1])\n",
        "\n",
        "#KNNs\n",
        "knn3 = KNeighborsClassifier(n_neighbors = 3)\n",
        "knn3.fit(Xtreino, Ytreino)\n",
        "pred_knn3 = knn3.predict(Xvalid)\n",
        "met_knn3 = MetricasClassBinaria(Yvalid, pred_knn3, custo = [0, 1, 10, 1])\n",
        "\n",
        "knn5 = KNeighborsClassifier(n_neighbors = 5)\n",
        "knn5.fit(Xtreino, Ytreino)\n",
        "pred_knn5 = knn5.predict(Xvalid)\n",
        "met_knn5 = MetricasClassBinaria(Yvalid, pred_knn5, custo = [0, 1, 10, 1])\n",
        "\n",
        "knn7 = KNeighborsClassifier(n_neighbors = 7)\n",
        "knn7.fit(Xtreino, Ytreino)\n",
        "pred_knn7 = knn7.predict(Xvalid)\n",
        "met_knn7 = MetricasClassBinaria(Yvalid, pred_knn7, custo = [0, 1, 10, 1])\n",
        "\n",
        "knn9 = KNeighborsClassifier(n_neighbors = 9)\n",
        "knn9.fit(Xtreino, Ytreino)\n",
        "pred_knn9 = knn9.predict(Xvalid)\n",
        "met_knn9 = MetricasClassBinaria(Yvalid, pred_knn9, custo = [0, 1, 10, 1])\n",
        "\n",
        "#Arvores\n",
        "arvore4 = DecisionTreeClassifier(max_depth = 4)\n",
        "arvore4.fit(Xtreino, Ytreino)\n",
        "pred_arvore4 = arvore4.predict(Xvalid)\n",
        "met_arvore4 = MetricasClassBinaria(Yvalid, pred_arvore4, custo = [0, 1, 10, 1])\n",
        "\n",
        "arvore8 = DecisionTreeClassifier(max_depth = 8)\n",
        "arvore8.fit(Xtreino, Ytreino)\n",
        "pred_arvore8 = arvore8.predict(Xvalid)\n",
        "met_arvore8 = MetricasClassBinaria(Yvalid, pred_arvore8, custo = [0, 1, 10, 1])\n",
        "\n",
        "arvore12 = DecisionTreeClassifier(max_depth = 12)\n",
        "arvore12.fit(Xtreino, Ytreino)\n",
        "pred_arvore12 = arvore12.predict(Xvalid)\n",
        "met_arvore12 = MetricasClassBinaria(Yvalid, pred_arvore12, custo = [0, 1, 10, 1])\n",
        "\n",
        "#Florestas\n",
        "rf8 = RandomForestClassifier(max_depth = 8)\n",
        "rf8.fit(Xtreino, Ytreino)\n",
        "pred_rf8 = rf8.predict(Xvalid)\n",
        "met_rf8 = MetricasClassBinaria(Yvalid, pred_rf8, custo = [0, 1, 10, 1])\n",
        "\n",
        "rf16 = RandomForestClassifier(max_depth = 16)\n",
        "rf16.fit(Xtreino, Ytreino)\n",
        "pred_rf16 = rf16.predict(Xvalid)\n",
        "met_rf16 = MetricasClassBinaria(Yvalid, pred_rf16, custo = [0, 1, 10, 1])\n",
        "\n",
        "rf24 = RandomForestClassifier(max_depth = 24)\n",
        "rf24.fit(Xtreino, Ytreino)\n",
        "pred_rf24 = rf24.predict(Xvalid)\n",
        "met_rf24 = MetricasClassBinaria(Yvalid, pred_rf24, custo = [0, 1, 10, 1])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8270\n",
            "             Sensibilidade: 0.3476\n",
            "            Especificidade: 0.9544\n",
            "  Valor Preditivo Positivo: 0.6697\n",
            "  Valor Preditivo Negativo: 0.8462\n",
            "Coef. de Corr. de Matthews: 0.3948\n",
            "                  F1-Score: 0.4577\n",
            "                Eficiência: 0.6510\n",
            "\n",
            "                     Custo: 1479\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8090\n",
            "             Sensibilidade: 0.3381\n",
            "            Especificidade: 0.9342\n",
            "  Valor Preditivo Positivo: 0.5772\n",
            "  Valor Preditivo Negativo: 0.8415\n",
            "Coef. de Corr. de Matthews: 0.3377\n",
            "                  F1-Score: 0.4264\n",
            "                Eficiência: 0.6361\n",
            "\n",
            "                     Custo: 1513\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8280\n",
            "             Sensibilidade: 0.3429\n",
            "            Especificidade: 0.9570\n",
            "  Valor Preditivo Positivo: 0.6792\n",
            "  Valor Preditivo Negativo: 0.8456\n",
            "Coef. de Corr. de Matthews: 0.3967\n",
            "                  F1-Score: 0.4557\n",
            "                Eficiência: 0.6499\n",
            "\n",
            "                     Custo: 1486\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8290\n",
            "             Sensibilidade: 0.3524\n",
            "            Especificidade: 0.9557\n",
            "  Valor Preditivo Positivo: 0.6789\n",
            "  Valor Preditivo Negativo: 0.8474\n",
            "Coef. de Corr. de Matthews: 0.4027\n",
            "                  F1-Score: 0.4639\n",
            "                Eficiência: 0.6540\n",
            "\n",
            "                     Custo: 1469\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8360\n",
            "             Sensibilidade: 0.3524\n",
            "            Especificidade: 0.9646\n",
            "  Valor Preditivo Positivo: 0.7255\n",
            "  Valor Preditivo Negativo: 0.8486\n",
            "Coef. de Corr. de Matthews: 0.4265\n",
            "                  F1-Score: 0.4744\n",
            "                Eficiência: 0.6585\n",
            "\n",
            "                     Custo: 1462\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8410\n",
            "             Sensibilidade: 0.3333\n",
            "            Especificidade: 0.9759\n",
            "  Valor Preditivo Positivo: 0.7865\n",
            "  Valor Preditivo Negativo: 0.8463\n",
            "Coef. de Corr. de Matthews: 0.4424\n",
            "                  F1-Score: 0.4682\n",
            "                Eficiência: 0.6546\n",
            "\n",
            "                     Custo: 1489\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8390\n",
            "             Sensibilidade: 0.4190\n",
            "            Especificidade: 0.9506\n",
            "  Valor Preditivo Positivo: 0.6929\n",
            "  Valor Preditivo Negativo: 0.8603\n",
            "Coef. de Corr. de Matthews: 0.4522\n",
            "                  F1-Score: 0.5223\n",
            "                Eficiência: 0.6848\n",
            "\n",
            "                     Custo: 1347\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8110\n",
            "             Sensibilidade: 0.4095\n",
            "            Especificidade: 0.9177\n",
            "  Valor Preditivo Positivo: 0.5695\n",
            "  Valor Preditivo Negativo: 0.8539\n",
            "Coef. de Corr. de Matthews: 0.3723\n",
            "                  F1-Score: 0.4765\n",
            "                Eficiência: 0.6636\n",
            "\n",
            "                     Custo: 1391\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8560\n",
            "             Sensibilidade: 0.3857\n",
            "            Especificidade: 0.9810\n",
            "  Valor Preditivo Positivo: 0.8438\n",
            "  Valor Preditivo Negativo: 0.8573\n",
            "Coef. de Corr. de Matthews: 0.5070\n",
            "                  F1-Score: 0.5294\n",
            "                Eficiência: 0.6834\n",
            "\n",
            "                     Custo: 1386\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8510\n",
            "             Sensibilidade: 0.4238\n",
            "            Especificidade: 0.9646\n",
            "  Valor Preditivo Positivo: 0.7607\n",
            "  Valor Preditivo Negativo: 0.8630\n",
            "Coef. de Corr. de Matthews: 0.4921\n",
            "                  F1-Score: 0.5443\n",
            "                Eficiência: 0.6942\n",
            "\n",
            "                     Custo: 1327\n",
            "  Prevalência de Positivos: 0.2100\n",
            "                  Acurácia: 0.8480\n",
            "             Sensibilidade: 0.4333\n",
            "            Especificidade: 0.9582\n",
            "  Valor Preditivo Positivo: 0.7339\n",
            "  Valor Preditivo Negativo: 0.8642\n",
            "Coef. de Corr. de Matthews: 0.4839\n",
            "                  F1-Score: 0.5449\n",
            "                Eficiência: 0.6958\n",
            "\n",
            "                     Custo: 1314\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yLh1cExMj9xB"
      },
      "source": [
        "modelos = ['Reg. Log.', '3-NN', '5-NN', '7-NN', '9-NN', \n",
        "           'Árvore (4)', 'Árvore (8)', 'Árvore (12)',\n",
        "           'Random Forest (8)', 'Random Forest (16)', 'Random Forest (24)']\n",
        "\n",
        "preditos = [met_reglog, met_knn3, met_knn5, met_knn7, met_knn9, met_arvore4,\n",
        "            met_arvore8, met_arvore12, met_rf8, met_rf16, met_rf24]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MpU24LAYjFnI"
      },
      "source": [
        "resultados = pd.DataFrame()\n",
        "\n",
        "for i in range(len(modelos)):\n",
        "    aux = pd.DataFrame.from_dict(preditos[i], orient = 'index', columns = [modelos[i]])\n",
        "    resultados = pd.concat([resultados, aux], axis = 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HcxaJm8dkw8w",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 394
        },
        "outputId": "fd1a2858-c5bd-4a29-a688-7874bbd69ab3"
      },
      "source": [
        "resultados.T.round(4)"
      ],
      "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>prevalência</th>\n",
              "      <th>acurácia</th>\n",
              "      <th>sensibilidade</th>\n",
              "      <th>especificidade</th>\n",
              "      <th>vpp</th>\n",
              "      <th>vpn</th>\n",
              "      <th>mcc</th>\n",
              "      <th>f1-score</th>\n",
              "      <th>eficiência</th>\n",
              "      <th>custo</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Reg. Log.</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.827</td>\n",
              "      <td>0.3476</td>\n",
              "      <td>0.9544</td>\n",
              "      <td>0.6697</td>\n",
              "      <td>0.8462</td>\n",
              "      <td>0.3948</td>\n",
              "      <td>0.4577</td>\n",
              "      <td>0.6510</td>\n",
              "      <td>1479.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.809</td>\n",
              "      <td>0.3381</td>\n",
              "      <td>0.9342</td>\n",
              "      <td>0.5772</td>\n",
              "      <td>0.8415</td>\n",
              "      <td>0.3377</td>\n",
              "      <td>0.4264</td>\n",
              "      <td>0.6361</td>\n",
              "      <td>1513.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.828</td>\n",
              "      <td>0.3429</td>\n",
              "      <td>0.9570</td>\n",
              "      <td>0.6792</td>\n",
              "      <td>0.8456</td>\n",
              "      <td>0.3967</td>\n",
              "      <td>0.4557</td>\n",
              "      <td>0.6499</td>\n",
              "      <td>1486.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.829</td>\n",
              "      <td>0.3524</td>\n",
              "      <td>0.9557</td>\n",
              "      <td>0.6789</td>\n",
              "      <td>0.8474</td>\n",
              "      <td>0.4027</td>\n",
              "      <td>0.4639</td>\n",
              "      <td>0.6540</td>\n",
              "      <td>1469.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.836</td>\n",
              "      <td>0.3524</td>\n",
              "      <td>0.9646</td>\n",
              "      <td>0.7255</td>\n",
              "      <td>0.8486</td>\n",
              "      <td>0.4265</td>\n",
              "      <td>0.4744</td>\n",
              "      <td>0.6585</td>\n",
              "      <td>1462.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (4)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.841</td>\n",
              "      <td>0.3333</td>\n",
              "      <td>0.9759</td>\n",
              "      <td>0.7865</td>\n",
              "      <td>0.8463</td>\n",
              "      <td>0.4424</td>\n",
              "      <td>0.4682</td>\n",
              "      <td>0.6546</td>\n",
              "      <td>1489.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (8)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.839</td>\n",
              "      <td>0.4190</td>\n",
              "      <td>0.9506</td>\n",
              "      <td>0.6929</td>\n",
              "      <td>0.8603</td>\n",
              "      <td>0.4522</td>\n",
              "      <td>0.5223</td>\n",
              "      <td>0.6848</td>\n",
              "      <td>1347.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (12)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.811</td>\n",
              "      <td>0.4095</td>\n",
              "      <td>0.9177</td>\n",
              "      <td>0.5695</td>\n",
              "      <td>0.8539</td>\n",
              "      <td>0.3723</td>\n",
              "      <td>0.4765</td>\n",
              "      <td>0.6636</td>\n",
              "      <td>1391.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (8)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.856</td>\n",
              "      <td>0.3857</td>\n",
              "      <td>0.9810</td>\n",
              "      <td>0.8438</td>\n",
              "      <td>0.8573</td>\n",
              "      <td>0.5070</td>\n",
              "      <td>0.5294</td>\n",
              "      <td>0.6834</td>\n",
              "      <td>1386.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (16)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.851</td>\n",
              "      <td>0.4238</td>\n",
              "      <td>0.9646</td>\n",
              "      <td>0.7607</td>\n",
              "      <td>0.8630</td>\n",
              "      <td>0.4921</td>\n",
              "      <td>0.5443</td>\n",
              "      <td>0.6942</td>\n",
              "      <td>1327.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (24)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.848</td>\n",
              "      <td>0.4333</td>\n",
              "      <td>0.9582</td>\n",
              "      <td>0.7339</td>\n",
              "      <td>0.8642</td>\n",
              "      <td>0.4839</td>\n",
              "      <td>0.5449</td>\n",
              "      <td>0.6958</td>\n",
              "      <td>1314.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                    prevalência  acurácia  ...  eficiência   custo\n",
              "Reg. Log.                  0.21     0.827  ...      0.6510  1479.0\n",
              "3-NN                       0.21     0.809  ...      0.6361  1513.0\n",
              "5-NN                       0.21     0.828  ...      0.6499  1486.0\n",
              "7-NN                       0.21     0.829  ...      0.6540  1469.0\n",
              "9-NN                       0.21     0.836  ...      0.6585  1462.0\n",
              "Árvore (4)                 0.21     0.841  ...      0.6546  1489.0\n",
              "Árvore (8)                 0.21     0.839  ...      0.6848  1347.0\n",
              "Árvore (12)                0.21     0.811  ...      0.6636  1391.0\n",
              "Random Forest (8)          0.21     0.856  ...      0.6834  1386.0\n",
              "Random Forest (16)         0.21     0.851  ...      0.6942  1327.0\n",
              "Random Forest (24)         0.21     0.848  ...      0.6958  1314.0\n",
              "\n",
              "[11 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 145
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bCEOh8vJk5vC"
      },
      "source": [
        "from google.colab.data_table import DataTable"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r9mwNUWfm7Hz",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 319
        },
        "outputId": "ce3f815a-1d14-4622-8cca-3f11879d6239"
      },
      "source": [
        "DataTable(resultados.T.round(4))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.module+javascript": "\n      import \"https://ssl.gstatic.com/colaboratory/data_table/a6224c040fa35dcf/data_table.js\";\n\n      window.createDataTable({\n        data: [[\"Reg. Log.\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.827,\n            'f': \"0.827\",\n        },\n{\n            'v': 0.3476,\n            'f': \"0.3476\",\n        },\n{\n            'v': 0.9544,\n            'f': \"0.9544\",\n        },\n{\n            'v': 0.6697,\n            'f': \"0.6697\",\n        },\n{\n            'v': 0.8462,\n            'f': \"0.8462\",\n        },\n{\n            'v': 0.3948,\n            'f': \"0.3948\",\n        },\n{\n            'v': 0.4577,\n            'f': \"0.4577\",\n        },\n{\n            'v': 0.651,\n            'f': \"0.651\",\n        },\n{\n            'v': 1479.0,\n            'f': \"1479.0\",\n        }],\n [\"3-NN\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.809,\n            'f': \"0.809\",\n        },\n{\n            'v': 0.3381,\n            'f': \"0.3381\",\n        },\n{\n            'v': 0.9342,\n            'f': \"0.9342\",\n        },\n{\n            'v': 0.5772,\n            'f': \"0.5772\",\n        },\n{\n            'v': 0.8415,\n            'f': \"0.8415\",\n        },\n{\n            'v': 0.3377,\n            'f': \"0.3377\",\n        },\n{\n            'v': 0.4264,\n            'f': \"0.4264\",\n        },\n{\n            'v': 0.6361,\n            'f': \"0.6361\",\n        },\n{\n            'v': 1513.0,\n            'f': \"1513.0\",\n        }],\n [\"5-NN\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.828,\n            'f': \"0.828\",\n        },\n{\n            'v': 0.3429,\n            'f': \"0.3429\",\n        },\n{\n            'v': 0.957,\n            'f': \"0.957\",\n        },\n{\n            'v': 0.6792,\n            'f': \"0.6792\",\n        },\n{\n            'v': 0.8456,\n            'f': \"0.8456\",\n        },\n{\n            'v': 0.3967,\n            'f': \"0.3967\",\n        },\n{\n            'v': 0.4557,\n            'f': \"0.4557\",\n        },\n{\n            'v': 0.6499,\n            'f': \"0.6499\",\n        },\n{\n            'v': 1486.0,\n            'f': \"1486.0\",\n        }],\n [\"7-NN\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.829,\n            'f': \"0.829\",\n        },\n{\n            'v': 0.3524,\n            'f': \"0.3524\",\n        },\n{\n            'v': 0.9557,\n            'f': \"0.9557\",\n        },\n{\n            'v': 0.6789,\n            'f': \"0.6789\",\n        },\n{\n            'v': 0.8474,\n            'f': \"0.8474\",\n        },\n{\n            'v': 0.4027,\n            'f': \"0.4027\",\n        },\n{\n            'v': 0.4639,\n            'f': \"0.4639\",\n        },\n{\n            'v': 0.654,\n            'f': \"0.654\",\n        },\n{\n            'v': 1469.0,\n            'f': \"1469.0\",\n        }],\n [\"9-NN\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.836,\n            'f': \"0.836\",\n        },\n{\n            'v': 0.3524,\n            'f': \"0.3524\",\n        },\n{\n            'v': 0.9646,\n            'f': \"0.9646\",\n        },\n{\n            'v': 0.7255,\n            'f': \"0.7255\",\n        },\n{\n            'v': 0.8486,\n            'f': \"0.8486\",\n        },\n{\n            'v': 0.4265,\n            'f': \"0.4265\",\n        },\n{\n            'v': 0.4744,\n            'f': \"0.4744\",\n        },\n{\n            'v': 0.6585,\n            'f': \"0.6585\",\n        },\n{\n            'v': 1462.0,\n            'f': \"1462.0\",\n        }],\n [\"\\u00c1rvore (4)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.841,\n            'f': \"0.841\",\n        },\n{\n            'v': 0.3333,\n            'f': \"0.3333\",\n        },\n{\n            'v': 0.9759,\n            'f': \"0.9759\",\n        },\n{\n            'v': 0.7865,\n            'f': \"0.7865\",\n        },\n{\n            'v': 0.8463,\n            'f': \"0.8463\",\n        },\n{\n            'v': 0.4424,\n            'f': \"0.4424\",\n        },\n{\n            'v': 0.4682,\n            'f': \"0.4682\",\n        },\n{\n            'v': 0.6546,\n            'f': \"0.6546\",\n        },\n{\n            'v': 1489.0,\n            'f': \"1489.0\",\n        }],\n [\"\\u00c1rvore (8)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.839,\n            'f': \"0.839\",\n        },\n{\n            'v': 0.419,\n            'f': \"0.419\",\n        },\n{\n            'v': 0.9506,\n            'f': \"0.9506\",\n        },\n{\n            'v': 0.6929,\n            'f': \"0.6929\",\n        },\n{\n            'v': 0.8603,\n            'f': \"0.8603\",\n        },\n{\n            'v': 0.4522,\n            'f': \"0.4522\",\n        },\n{\n            'v': 0.5223,\n            'f': \"0.5223\",\n        },\n{\n            'v': 0.6848,\n            'f': \"0.6848\",\n        },\n{\n            'v': 1347.0,\n            'f': \"1347.0\",\n        }],\n [\"\\u00c1rvore (12)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.811,\n            'f': \"0.811\",\n        },\n{\n            'v': 0.4095,\n            'f': \"0.4095\",\n        },\n{\n            'v': 0.9177,\n            'f': \"0.9177\",\n        },\n{\n            'v': 0.5695,\n            'f': \"0.5695\",\n        },\n{\n            'v': 0.8539,\n            'f': \"0.8539\",\n        },\n{\n            'v': 0.3723,\n            'f': \"0.3723\",\n        },\n{\n            'v': 0.4765,\n            'f': \"0.4765\",\n        },\n{\n            'v': 0.6636,\n            'f': \"0.6636\",\n        },\n{\n            'v': 1391.0,\n            'f': \"1391.0\",\n        }],\n [\"Random Forest (8)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.856,\n            'f': \"0.856\",\n        },\n{\n            'v': 0.3857,\n            'f': \"0.3857\",\n        },\n{\n            'v': 0.981,\n            'f': \"0.981\",\n        },\n{\n            'v': 0.8438,\n            'f': \"0.8438\",\n        },\n{\n            'v': 0.8573,\n            'f': \"0.8573\",\n        },\n{\n            'v': 0.507,\n            'f': \"0.507\",\n        },\n{\n            'v': 0.5294,\n            'f': \"0.5294\",\n        },\n{\n            'v': 0.6834,\n            'f': \"0.6834\",\n        },\n{\n            'v': 1386.0,\n            'f': \"1386.0\",\n        }],\n [\"Random Forest (16)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.851,\n            'f': \"0.851\",\n        },\n{\n            'v': 0.4238,\n            'f': \"0.4238\",\n        },\n{\n            'v': 0.9646,\n            'f': \"0.9646\",\n        },\n{\n            'v': 0.7607,\n            'f': \"0.7607\",\n        },\n{\n            'v': 0.863,\n            'f': \"0.863\",\n        },\n{\n            'v': 0.4921,\n            'f': \"0.4921\",\n        },\n{\n            'v': 0.5443,\n            'f': \"0.5443\",\n        },\n{\n            'v': 0.6942,\n            'f': \"0.6942\",\n        },\n{\n            'v': 1327.0,\n            'f': \"1327.0\",\n        }],\n [\"Random Forest (24)\",\n{\n            'v': 0.21,\n            'f': \"0.21\",\n        },\n{\n            'v': 0.848,\n            'f': \"0.848\",\n        },\n{\n            'v': 0.4333,\n            'f': \"0.4333\",\n        },\n{\n            'v': 0.9582,\n            'f': \"0.9582\",\n        },\n{\n            'v': 0.7339,\n            'f': \"0.7339\",\n        },\n{\n            'v': 0.8642,\n            'f': \"0.8642\",\n        },\n{\n            'v': 0.4839,\n            'f': \"0.4839\",\n        },\n{\n            'v': 0.5449,\n            'f': \"0.5449\",\n        },\n{\n            'v': 0.6958,\n            'f': \"0.6958\",\n        },\n{\n            'v': 1314.0,\n            'f': \"1314.0\",\n        }]],\n        columns: [[\"string\", \"index\"], [\"number\", \"preval\\u00eancia\"], [\"number\", \"acur\\u00e1cia\"], [\"number\", \"sensibilidade\"], [\"number\", \"especificidade\"], [\"number\", \"vpp\"], [\"number\", \"vpn\"], [\"number\", \"mcc\"], [\"number\", \"f1-score\"], [\"number\", \"efici\\u00eancia\"], [\"number\", \"custo\"]],\n        columnOptions: [{\"width\": \"1px\", \"className\": \"index_column\"}],\n        rowsPerPage: 25,\n        helpUrl: \"https://colab.research.google.com/notebooks/data_table.ipynb\",\n        suppressOutputScrolling: true,\n        minimumWidth: undefined,\n      });\n    ",
            "text/plain": [
              "<google.colab.data_table.DataTable object>"
            ],
            "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>prevalência</th>\n",
              "      <th>acurácia</th>\n",
              "      <th>sensibilidade</th>\n",
              "      <th>especificidade</th>\n",
              "      <th>vpp</th>\n",
              "      <th>vpn</th>\n",
              "      <th>mcc</th>\n",
              "      <th>f1-score</th>\n",
              "      <th>eficiência</th>\n",
              "      <th>custo</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Reg. Log.</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.827</td>\n",
              "      <td>0.3476</td>\n",
              "      <td>0.9544</td>\n",
              "      <td>0.6697</td>\n",
              "      <td>0.8462</td>\n",
              "      <td>0.3948</td>\n",
              "      <td>0.4577</td>\n",
              "      <td>0.6510</td>\n",
              "      <td>1479.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.809</td>\n",
              "      <td>0.3381</td>\n",
              "      <td>0.9342</td>\n",
              "      <td>0.5772</td>\n",
              "      <td>0.8415</td>\n",
              "      <td>0.3377</td>\n",
              "      <td>0.4264</td>\n",
              "      <td>0.6361</td>\n",
              "      <td>1513.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.828</td>\n",
              "      <td>0.3429</td>\n",
              "      <td>0.9570</td>\n",
              "      <td>0.6792</td>\n",
              "      <td>0.8456</td>\n",
              "      <td>0.3967</td>\n",
              "      <td>0.4557</td>\n",
              "      <td>0.6499</td>\n",
              "      <td>1486.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.829</td>\n",
              "      <td>0.3524</td>\n",
              "      <td>0.9557</td>\n",
              "      <td>0.6789</td>\n",
              "      <td>0.8474</td>\n",
              "      <td>0.4027</td>\n",
              "      <td>0.4639</td>\n",
              "      <td>0.6540</td>\n",
              "      <td>1469.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9-NN</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.836</td>\n",
              "      <td>0.3524</td>\n",
              "      <td>0.9646</td>\n",
              "      <td>0.7255</td>\n",
              "      <td>0.8486</td>\n",
              "      <td>0.4265</td>\n",
              "      <td>0.4744</td>\n",
              "      <td>0.6585</td>\n",
              "      <td>1462.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (4)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.841</td>\n",
              "      <td>0.3333</td>\n",
              "      <td>0.9759</td>\n",
              "      <td>0.7865</td>\n",
              "      <td>0.8463</td>\n",
              "      <td>0.4424</td>\n",
              "      <td>0.4682</td>\n",
              "      <td>0.6546</td>\n",
              "      <td>1489.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (8)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.839</td>\n",
              "      <td>0.4190</td>\n",
              "      <td>0.9506</td>\n",
              "      <td>0.6929</td>\n",
              "      <td>0.8603</td>\n",
              "      <td>0.4522</td>\n",
              "      <td>0.5223</td>\n",
              "      <td>0.6848</td>\n",
              "      <td>1347.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Árvore (12)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.811</td>\n",
              "      <td>0.4095</td>\n",
              "      <td>0.9177</td>\n",
              "      <td>0.5695</td>\n",
              "      <td>0.8539</td>\n",
              "      <td>0.3723</td>\n",
              "      <td>0.4765</td>\n",
              "      <td>0.6636</td>\n",
              "      <td>1391.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (8)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.856</td>\n",
              "      <td>0.3857</td>\n",
              "      <td>0.9810</td>\n",
              "      <td>0.8438</td>\n",
              "      <td>0.8573</td>\n",
              "      <td>0.5070</td>\n",
              "      <td>0.5294</td>\n",
              "      <td>0.6834</td>\n",
              "      <td>1386.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (16)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.851</td>\n",
              "      <td>0.4238</td>\n",
              "      <td>0.9646</td>\n",
              "      <td>0.7607</td>\n",
              "      <td>0.8630</td>\n",
              "      <td>0.4921</td>\n",
              "      <td>0.5443</td>\n",
              "      <td>0.6942</td>\n",
              "      <td>1327.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Random Forest (24)</th>\n",
              "      <td>0.21</td>\n",
              "      <td>0.848</td>\n",
              "      <td>0.4333</td>\n",
              "      <td>0.9582</td>\n",
              "      <td>0.7339</td>\n",
              "      <td>0.8642</td>\n",
              "      <td>0.4839</td>\n",
              "      <td>0.5449</td>\n",
              "      <td>0.6958</td>\n",
              "      <td>1314.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 147
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6aATBoI5khr0"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gJTzEY5bq9HX"
      },
      "source": [
        "### Quais as variáveis que são mais relevantes nesse problema?"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ouOK3nAmDn09",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "outputId": "0c07dc9e-1865-4829-c7e7-d0b8a439fd86"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "nomes = list(Xtreino.columns)\n",
        "x = np.arange(len(nomes))\n",
        "y = rf24.feature_importances_\n",
        " \n",
        "plt.style.use('ggplot')\n",
        "data = pd.DataFrame({'Importância': y}, index = nomes) \n",
        "data.sort_values(by = 'Importância').plot(kind = 'barh', colormap = 'Dark2')\n",
        "plt.title('Gráfico de importância das variáveis')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZLsPBMXkkgfJ"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qx104lLyrBrf"
      },
      "source": [
        "### Enfim, qual estratégia tomar agora?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kltqJ3AjPAN8"
      },
      "source": [
        "De acordo com nossos resultados preliminares, temos que o modelo baseado em Árvore de Decisão, com profundidade igual 12, retorna o menor custo na amostra de validação. \n",
        "\n",
        "Vamos utilizar o método predict_proba para prever as probabilidades de cada indivíduo na amostra de validação entrar em churn e classificá-los em faixas.\n",
        "\n",
        "Podemos compor as faixas da seguinte maneira:\n",
        "\n",
        "- 0% - 25% de probabilidade de Churn: Ações de Marketing Leves\n",
        "- 25% - 50% de probabilidade de Churn: Ações de Marketing Intermediárias\n",
        "- 50% - 75% de probabilidade de Churn: Ações de Marketing Fortes\n",
        "- 75% - 100% de probabilidade de Churn: Ações de Marketing Agressivas"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NtIMkhurQCnM"
      },
      "source": [
        "probs = pd.DataFrame(rf24.predict_proba(Xvalid))\n",
        "\n",
        "def aux(x):\n",
        "    if x < 0.25:\n",
        "        return 'Leve'\n",
        "    elif x < 0.5:\n",
        "        return 'Intermediário'\n",
        "    elif x < 0.75:\n",
        "        return 'Forte'\n",
        "    else:\n",
        "        return 'Agressivo'\n",
        "\n",
        "probs['Campanha'] = probs[1].apply(aux)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uWvFX4uV9MZg",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 676
        },
        "outputId": "5f811513-b634-45c0-bb61-d5923a5e4e3e"
      },
      "source": [
        "probs.sample(20)"
      ],
      "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>0</th>\n",
              "      <th>1</th>\n",
              "      <th>Campanha</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>714</th>\n",
              "      <td>0.950000</td>\n",
              "      <td>0.050000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>87</th>\n",
              "      <td>0.630000</td>\n",
              "      <td>0.370000</td>\n",
              "      <td>Intermediário</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>603</th>\n",
              "      <td>0.910000</td>\n",
              "      <td>0.090000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>894</th>\n",
              "      <td>0.917500</td>\n",
              "      <td>0.082500</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>185</th>\n",
              "      <td>0.830000</td>\n",
              "      <td>0.170000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>718</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>270</th>\n",
              "      <td>0.210000</td>\n",
              "      <td>0.790000</td>\n",
              "      <td>Agressivo</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>261</th>\n",
              "      <td>0.827500</td>\n",
              "      <td>0.172500</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>414</th>\n",
              "      <td>0.990000</td>\n",
              "      <td>0.010000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>549</th>\n",
              "      <td>0.930000</td>\n",
              "      <td>0.070000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>867</th>\n",
              "      <td>0.930000</td>\n",
              "      <td>0.070000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>97</th>\n",
              "      <td>0.960000</td>\n",
              "      <td>0.040000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498</th>\n",
              "      <td>0.936667</td>\n",
              "      <td>0.063333</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>331</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>541</th>\n",
              "      <td>0.830000</td>\n",
              "      <td>0.170000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>37</th>\n",
              "      <td>0.990000</td>\n",
              "      <td>0.010000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>622</th>\n",
              "      <td>0.960000</td>\n",
              "      <td>0.040000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>204</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>909</th>\n",
              "      <td>0.280000</td>\n",
              "      <td>0.720000</td>\n",
              "      <td>Forte</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>916</th>\n",
              "      <td>1.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>Leve</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "            0         1       Campanha\n",
              "714  0.950000  0.050000           Leve\n",
              "87   0.630000  0.370000  Intermediário\n",
              "603  0.910000  0.090000           Leve\n",
              "894  0.917500  0.082500           Leve\n",
              "185  0.830000  0.170000           Leve\n",
              "718  1.000000  0.000000           Leve\n",
              "270  0.210000  0.790000      Agressivo\n",
              "261  0.827500  0.172500           Leve\n",
              "414  0.990000  0.010000           Leve\n",
              "549  0.930000  0.070000           Leve\n",
              "867  0.930000  0.070000           Leve\n",
              "97   0.960000  0.040000           Leve\n",
              "498  0.936667  0.063333           Leve\n",
              "331  1.000000  0.000000           Leve\n",
              "541  0.830000  0.170000           Leve\n",
              "37   0.990000  0.010000           Leve\n",
              "622  0.960000  0.040000           Leve\n",
              "204  1.000000  0.000000           Leve\n",
              "909  0.280000  0.720000          Forte\n",
              "916  1.000000  0.000000           Leve"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 152
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GpPqh2GqRptC",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "44dec8ad-5325-4148-e754-0cba61d0d4db"
      },
      "source": [
        "probs['Campanha'].value_counts()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Leve             717\n",
              "Intermediário    157\n",
              "Forte             69\n",
              "Agressivo         57\n",
              "Name: Campanha, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 153
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sqX1oLPkQpzd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "b50c09d5-7cf4-4abe-adb7-b04834b17695"
      },
      "source": [
        "probs['Campanha'].value_counts().sort_values(ascending = True).plot(kind = 'barh')\n",
        "plt.title('Quantidade de clientes em cada ação a ser tomada')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zPrsaN6Lkejk"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rqG-vQ-mWIsz"
      },
      "source": [
        "# Criando uma Métrica Personalizada no Scikit-learn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YPajsZVqMIPN"
      },
      "source": [
        "#testando minha metrica\n",
        "def custo(observado, predito):\n",
        "    mc = confusion_matrix(observado, predito)\n",
        "    mcusto = [0, 1, 10, 1]\n",
        "    return (mc * np.array(mcusto).reshape(2,2)).sum()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "H4jZ5TftM7Vs",
        "outputId": "5526e0bc-f626-4e6e-cde3-bd29dd3d9f9d"
      },
      "source": [
        "from sklearn.metrics import make_scorer\n",
        "minhametrica = make_scorer(custo, greater_is_better = False)\n",
        "rl = cross_validate(LogisticRegression(), Xtreino, Ytreino, cv = kf, scoring = minhametrica)\n",
        "rl['test_score'].mean()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "-1227.7"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 157
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ynre_vmzWUmb"
      },
      "source": [
        "---"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gHnoQ9N7R8Dd"
      },
      "source": [
        "## **Exercício Final**\n",
        "\n",
        "Faça um projeto com suas ideias e adicione ao seu portfólio!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-nBEEw-Ojy2E"
      },
      "source": [
        "## **Links Úteis**\n",
        "\n",
        "- [Conjunto de dados de Churn no kaggle](https://www.kaggle.com/shubh0799/churn-modelling)"
      ]
    }
  ]
}