Notebooks/3_modelos_machine_learning.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "3_modelos_machine_learning.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOmcgrwGwk1zpIXoH4va8gP",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/EltonGS/covid-19_previsao_leito_UTI/blob/main/Notebooks/3_modelos_machine_learning.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oZ3Ms8HqfBXk"
      },
      "source": [
        "# Introdução\n",
        "\n",
        "Esse é o terceiro, e último, notebook desse projeto. Nele a base de dados disponibilizada pelo Hospital Sírio Libanês é utilizada para a o treinamento e teste de **Modelos de Macine Learning** com o objetivo de prever se um paciente precisará, ou não, ser admitido na UTI com base nos dados disponíveis. A base de dados foi tratada no notebook **tratamento_dados** e analisada no notebook **analise_dos_dados**.\n",
        "\n",
        "### Mas o que é Machine Learning ou Aprendizado de Máquina?\n",
        "\n",
        "De acordo com a **IBM**, **Machine Learning é o ramo da ciência da computação e inteligência artificial que foca no uso de dados e algoritmos para imitar a maneira como humanos pensam, melhorando sua acurácia gradualmente**.\n",
        "\n",
        "**Machine Learning** é um componente importante no crescente campo da **Ciência de Dados**. Através de métodos estatísticos, algoritmos são treinados para realizar classificação ou previsões, realizando descobertas em projetos de mineração de dados. Essas descobertas guiam tomadas de decisão de negócios ou em aplicações.\n",
        "\n",
        "### O uso de Machine Learning em Sistemas de Saúde\n",
        "\n",
        "<p style=\"clear:both;margin-bottom:20px;\"><a href=\"https://healthinformatics.uic.edu/blog/machine-learning-in-healthcare/\" rel=\"noreferrer\" target=\"_blank\"><img src=\"https://s3.amazonaws.com/utep-uploads/wp-content/uploads/UIC/2020/11/13094738/How-Machine-Learning-is-Applied-in-Healthcare-1.jpg\" alt=\"Six ways in which machine learning is used to improve healthcare.\" style=\"max-width:100%;\" /></a></p><p style=\"clear:both;margin-bottom:20px;\"><a href=\"https://healthinformatics.uic.edu\" rel=\"noreferrer\" target=\"_blank\">University of Illinois at Chicago </a></p>\n",
        "\n",
        "\n",
        "Conforme demonstrado pela imagem acima, da Universidade de Illinois em Chicago, Machine Learning já é utilizado em sistemas de saúde e ajudam na evolução de tratamentos.\n",
        "\n",
        "Para a base de dados disponibilizada pelo Hospital Sírio Libanês, o **objetivo é identificar se um paciente necessitará de tratamento na UTI com base nos dados informados**. É necessário **CLASSIFICAR** se o paciente precisa ou não de UTI.\n",
        "\n",
        "Para isso serão utilizados algoritmos de classificação.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X62OpXafoM_S"
      },
      "source": [
        "# Informações sobre os dados\n",
        "\n",
        "O **Hospital Sírio Libanês** disponibilizou a base de dados no **Kaggle** para a comunidade analisar e criar modelos de **Machine Learning** que ajudem a identificar a necessidade de **UTI** para os pacientes que chegam ao hospital com Covid-19.\n",
        "\n",
        "No notebook de tratamendo dos dados tem a explicação fornecida pelo Sírio Libanês em como trabalhar com os dados. Relembrando:\n",
        "\n",
        "- **O objetivo é identificar quais pacientes precisarão de UTI e também quais não precisarão**.\n",
        "\n",
        "- **A partir do momento que o paciente é admitido na UTI os dados não devem ser utilizados para treinar o modelo**.\n",
        "\n",
        "- **Clinicamente é interessante que o diagnóstico da necessidade de UTI, ou não, seja realizado o mais rapidamente possível**.\n",
        "\n",
        "Com base nessas especificações do Hospital Sírio Libanês a seguinte decisão foi tomada:\n",
        "\n",
        "- **Serão analisados os dados da primeira (entre 0 e 2 horas da chegada do paciente ao hospital). Tendo assim um modelo que pode ser utilizado assim que o paciente chega ao hospital**."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nFGxISmIzW28"
      },
      "source": [
        "# Modelos de Machine Learning utilizados\n",
        "\n",
        "O problema apresentado pelo Hospital Sírio Libanês é de classificação.\n",
        "\n",
        "Os modelos de classificação serão importados da biblioteca **[SciKit Learn](https://scikit-learn.org/stable/)**.\n",
        "\n",
        "Os **modelos escolhidos** para serem testados com a base de dados são:\n",
        "\n",
        "- **Random Forest Classifier**;\n",
        "\n",
        "- **Logistic Regression**;\n",
        "\n",
        "Os modelos do tipo **SVM (Support Vector Machine)**:\n",
        "\n",
        "* **SVC (Support Vector Classifier)**;\n",
        "\n",
        "* **LinearSVC ( Linear Support Vector Classifier)**;\n",
        "\n",
        "Os modelos serão explicados quando utilizados.\n",
        "\n",
        "### Separação dos dados entre treino e teste\n",
        "\n",
        "Para o treinamento e teste do modelo é necessário a separação da base dados entre treino e teste. Para esse projeto a separação será 80% dos dados para treino e 20% dos dados para teste.\n",
        "\n",
        "Essa divisão será realizada utilizando os metódos **RepeatedStratifiedKFold** e **train_test_split** disponibilizados pelo **SciKit Learn**.\n",
        "\n",
        "### Métricas de avaliação\n",
        "\n",
        "As **métricas de avaliação** utilizadas para medir a **eficácia** dos modelos de Machine Learning serão:\n",
        "\n",
        "- **ROC AUC (Receiver Operating Characteristic - Area Under the Curve)**: A curva ROC demonstra o quão bom o modelo é em definir entre dois valores, por isso é utilizado para modelos de classficação. A métrica ROC possui 2 parâmetros:\n",
        "\n",
        "> * **Taxa de verdadeiro positivo**: que é medida por *verdadeiro positivo* / *(verdadeiro positivo + falso negativos)*;\n",
        "\n",
        "> * **Taxa de falso positivo**: é medida por *falso positivo* / *(falso positivo + verdadeiro negativo)*\n",
        "\n",
        "O **AUC** é a área sob a curva e varia entre 0 e 1. Quanto mais próximo de 1 melhor está o modelo.\n",
        "\n",
        "- **Matriz de Confusão (Confusion Matrix)**: é uma medida de performance para um modelo de classificação. É uma tabela com 4 diferentes combinações dos valores reais e os previstos. O resultado é uma matriz com 4 valores:\n",
        "\n",
        "> - **Verdadeiro Positivo**: o valor real e o previsto são positivos;\n",
        "> - **Verdadeiro Negativo**: o valor real e o previsto são negativos;\n",
        "> - **Falso Positivo**: o valor real é negativo e foi previsto como positivo;\n",
        "> - **Falso Negativo**: o valor real é positivo e foi previsto como negativo;\n",
        "\n",
        "Após a previsão realizada pelo modelo é possível extrair outras métricas como *Precision, Recall, F1 Score*.\n",
        "\n",
        "- **Precision**: de todas os valores previstos como positivos, quantos eram positivos no dados reais. Utilizado quando o falso positivo tem um resultado muito impactante quando comparado com o falso negativo;\n",
        "- **Recall**: de todos os valores positivos, quantos foram os previstos como positivos. Utilizado quando o falso negativo tem um impacto muito maior que o falso positivo;\n",
        "- **F1 Score**: média harmônica entre o o Precision e o Recall."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "StVQdhrOjpzj"
      },
      "source": [
        "# Recursos Utilizados\n",
        "\n",
        "Essa seção contém as **bibliotecas** e **funções** utilizadas nesse notebook.\n",
        "\n",
        "### Bibliotecas utilizadas"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A0f7mzjJe-GL"
      },
      "source": [
        "# manipulação dos dados\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "\n",
        "# visualização dos dados\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "\n",
        "# modelos de machine learning\n",
        "from sklearn.ensemble import RandomForestClassifier\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.svm import SVC, LinearSVC\n",
        "\n",
        "# divisão dos dados\n",
        "from sklearn.model_selection import cross_validate, RepeatedStratifiedKFold\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "# teste dos hiperparâmetros\n",
        "from sklearn.model_selection import RandomizedSearchCV\n",
        "\n",
        "# métricas de avaliação do modelo\n",
        "from sklearn.metrics import roc_curve, roc_auc_score, plot_roc_curve\n",
        "from sklearn.metrics import plot_confusion_matrix, classification_report"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oYgoLD6Cn3-r"
      },
      "source": [
        "### Funções"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uOmrDFDBn55W"
      },
      "source": [
        "def dimensao_dados (dados):\n",
        "  \"\"\"\n",
        "  Essa função verifica as dimensões do dataframe\n",
        "  \n",
        "  Retorna a quantidade de linhas e colunas\n",
        "  \"\"\"\n",
        "  # verifica as dimensões do dataframe\n",
        "  linha, coluna = dados.shape\n",
        "\n",
        "  # imprime as dimensões do dataframe\n",
        "  return print(f'O dataframe possui {linha} linhas e {coluna} colunas.')"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S8FI5h8WvlIs"
      },
      "source": [
        "def divide_dados(dados):\n",
        "  \"\"\"\n",
        "  Função para dividir os dados em treino e teste\n",
        "  ===============================================\n",
        "  Retorna os valores de X e y para treino e teste\n",
        "  \"\"\"\n",
        "  # separa os dados\n",
        "  x_colunas = dados.describe().columns\n",
        "  y = dados['icu']\n",
        "  x = dados[x_colunas].drop(['icu'], axis=1)\n",
        "\n",
        "  # divide os dados em treino e teste\n",
        "  x_train, x_test, y_train, y_test = train_test_split(x, y, stratify = y,\n",
        "                                                      random_state=481516, test_size=0.2)\n",
        "  \n",
        "  return x_train, x_test, y_train, y_test"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PlER2aq2p6j9"
      },
      "source": [
        "def cv_modelo(modelo, dados, n_splits, n_repeats):\n",
        "  \"\"\"\n",
        "  Executa a validação cruzada para o modelo informado.\n",
        "\n",
        "  Retorna o auc_medio para os dados de treino e teste\n",
        "  \"\"\"\n",
        "  \n",
        "  # define a seed\n",
        "  np.random.seed(481516)\n",
        "\n",
        "  # embaralha os dados\n",
        "  dados = dados.sample(frac=1).reset_index(drop=True) # frac = 1 pega todos os dados\n",
        "\n",
        "  # define as colunas a serem utilizadas\n",
        "  x_columns = dados.columns\n",
        "  y = dados['icu']\n",
        "  x = dados[x_columns].drop(['icu', 'window'], axis=1)\n",
        "\n",
        "  # instancia o Repeated Stratified K Fold\n",
        "  cv = RepeatedStratifiedKFold(n_splits=n_splits, n_repeats=n_repeats)\n",
        "\n",
        "  # realiza a validação cruzada\n",
        "  resultados = cross_validate(modelo, x, y, cv=cv, scoring='roc_auc', return_train_score=True) # return_train_score retorna os testes realizados com os dados de treino,\n",
        "                                                                                              # utilizado para visualizar o max_depth da decision tree\n",
        "\n",
        "  auc_medio = np.mean(resultados['test_score'])\n",
        "  auc_medio_treino = np.mean(resultados['train_score'])\n",
        "  auc_std = np.std(resultados['test_score'])\n",
        "\n",
        "  print(f'AUC médio para os dados de teste: {auc_medio}')\n",
        "  print(f'AUC médio para os dados de treino: {auc_medio_treino}')\n",
        "  print(f'Intervalo da AUC para os dados de teste: {auc_medio - (2* auc_std)} - {auc_medio + (2* auc_std)}')\n",
        "\n",
        "  return auc_medio, auc_medio_treino"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XESOa-Qi1ykA"
      },
      "source": [
        "def teste_hiperparametro(modelo, dados, hiperparametros):\n",
        "  \"\"\"\n",
        "  Função que testa hiperparâmetro para um determinado modelo.\n",
        "\n",
        "  Retorna os melhores hiperparâmetros\n",
        "  \"\"\"\n",
        "  # define a seed\n",
        "  np.random.seed(481516)\n",
        "\n",
        "  # embaralha os dados\n",
        "  dados = dados.sample(frac=1).reset_index(drop=True) # frac = 1 pega todos os dados\n",
        "\n",
        "  # define as colunas a serem utilizadas\n",
        "  x_columns = dados.columns\n",
        "  y = dados['icu']\n",
        "  x = dados[x_columns].drop(['icu', 'window'], axis=1)\n",
        "\n",
        "  # divisão dos dados\n",
        "  cv = RepeatedStratifiedKFold(random_state=481516)\n",
        "\n",
        "  # testa os hiperparâmetros\n",
        "  modelo_hiper_teste = RandomizedSearchCV(modelo, param_distributions=hiperparametros,\n",
        "                                          cv=cv, scoring='roc_auc', random_state=481516,\n",
        "                                          n_jobs=-1)\n",
        "  \n",
        "  hiper_param_selec = modelo_hiper_teste.fit(x, y)\n",
        "\n",
        "  return hiper_param_selec.best_params_, hiper_param_selec.best_score_"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eTMNQQNKuVdF"
      },
      "source": [
        "def plota_curva_roc(modelo, tipo_modelo, dados, cor_curva):\n",
        "  \"\"\"\n",
        "  Plota um gráfico com a curva ROC para o modelo espeficicado.\n",
        "\n",
        "  ============================================================\n",
        "\n",
        "  Parâmetros:\n",
        "\n",
        "  modelo: modelo instanciado\n",
        "\n",
        "  tipo_modelo: qual o modelo que será plotado\n",
        "\n",
        "  dados: dataframe utilizado no modelo\n",
        "\n",
        "  cor_cuva: cor da curva que será plotada\n",
        "  \"\"\"\n",
        "  \n",
        "  # divide os dados\n",
        "  x_train, x_test, y_train, y_test = divide_dados(dados)\n",
        "  \n",
        "  # cria a figura\n",
        "  fig = plt.subplots(figsize=(10, 10))\n",
        "  ax = plt.axes()\n",
        "\n",
        "  # plota a curva AUC\n",
        "  plot_roc_curve(modelo, x_test, y_test, ax=ax, color=cor_curva)\n",
        "\n",
        "  # define o título e o labels para os eixos\n",
        "  plt.title(f'Curva ROC para o modelo {tipo_modelo}\\n', fontsize=20, fontweight='bold',\n",
        "           color='black')\n",
        "  plt.ylabel('Razão de Verdadeiro Positivo (VP)', fontsize=11, color='dimgrey',\n",
        "             fontweight='bold')\n",
        "  plt.xlabel('Razão de Falto Positivo (FP)', fontsize=11, color='dimgrey',\n",
        "             fontweight='bold')\n",
        "\n",
        "  # parâmetros das linhas ao redor do gráfico\n",
        "  ax.spines['bottom'].set_color('silver')\n",
        "  ax.spines['left'].set_color('silver')\n",
        "  ax.spines['right'].set_visible(False)\n",
        "  ax.spines['top'].set_visible(False)\n",
        "\n",
        "  # cria a grid\n",
        "  ax.grid(axis='both', which='major', linestyle='dotted')\n",
        "\n",
        "  # manipula a legenda\n",
        "  ax.legend(loc='upper left', fontsize=10)\n",
        "\n",
        "  plt.show()"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6I-5GJianQsF"
      },
      "source": [
        "def plota_matriz_confusa(modelo, tipo_modelo, dados, paleta_matriz):\n",
        "  \"\"\"\n",
        "  Função plota a matriz de confusão do modelo informado\n",
        "\n",
        "  =====================================================\n",
        "\n",
        "  Parâmetros:\n",
        "\n",
        "  modelo: modelo instanciado\n",
        "\n",
        "  dados: dataframe utilizado para treinar/testar o modelo\n",
        "\n",
        "  paleta_matriz: paleta de cores utilizada para plotar a matriz\n",
        "  \"\"\"\n",
        "\n",
        "  # divide os dados\n",
        "  x_train, x_test, y_train, y_test = divide_dados(dados)\n",
        "\n",
        "  # cria a figura e os eixos\n",
        "  fig = plt.figure(figsize=(10, 10))\n",
        "  ax = plt.axes()\n",
        "\n",
        "  # plota a matriz\n",
        "  plot_confusion_matrix(modelo, x_test, y_test, ax = ax, cmap = paleta_matriz,\n",
        "                        display_labels=['UTI NÃO', 'UTI SIM'])\n",
        "  \n",
        "  # remove a colorbar\n",
        "  fig.delaxes(fig.axes[1])\n",
        "  \n",
        "  # define o título e o labels para os eixos\n",
        "  plt.title(f'Matriz de Confusão para o modelo {tipo_modelo}\\n', fontsize=20, fontweight='bold',\n",
        "           color='black')\n",
        "  plt.ylabel('CLASSIFICAÇÃO REAL', fontsize=11, color='dimgrey',\n",
        "             fontweight='bold')\n",
        "  plt.xlabel('CLASSIFICAÇÃO PREVISTA', fontsize=11, color='dimgrey',\n",
        "             fontweight='bold')\n",
        "  \n",
        "  plt.show()"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lPk0QZs_nVtd"
      },
      "source": [
        "# Carregando os dados"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "6GZ5hqIpmlOk",
        "outputId": "c198746b-c949-478c-e4bc-dfa4150b6c39"
      },
      "source": [
        "# endereço dos dados\n",
        "url = 'https://raw.githubusercontent.com/EltonGS/covid_ML_previsao_UTI/main/Dados/Dados%20Tratados/dados_ml.csv'\n",
        "\n",
        "# carrega os dados\n",
        "dados = pd.read_csv(url)\n",
        "\n",
        "# visualiza os dados\n",
        "dados.head()"
      ],
      "execution_count": 8,
      "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>patient_visit_identifier</th>\n",
              "      <th>age_above65</th>\n",
              "      <th>age_percentil</th>\n",
              "      <th>gender</th>\n",
              "      <th>disease grouping 1</th>\n",
              "      <th>disease grouping 2</th>\n",
              "      <th>disease grouping 3</th>\n",
              "      <th>disease grouping 4</th>\n",
              "      <th>disease grouping 5</th>\n",
              "      <th>disease grouping 6</th>\n",
              "      <th>htn</th>\n",
              "      <th>immunocompromised</th>\n",
              "      <th>other</th>\n",
              "      <th>albumin_median</th>\n",
              "      <th>albumin_diff</th>\n",
              "      <th>be_arterial_median</th>\n",
              "      <th>be_arterial_diff</th>\n",
              "      <th>be_venous_median</th>\n",
              "      <th>be_venous_diff</th>\n",
              "      <th>bic_arterial_median</th>\n",
              "      <th>bic_arterial_diff</th>\n",
              "      <th>bic_venous_median</th>\n",
              "      <th>bic_venous_diff</th>\n",
              "      <th>billirubin_median</th>\n",
              "      <th>billirubin_diff</th>\n",
              "      <th>blast_median</th>\n",
              "      <th>blast_diff</th>\n",
              "      <th>calcium_median</th>\n",
              "      <th>calcium_diff</th>\n",
              "      <th>creatinin_median</th>\n",
              "      <th>creatinin_diff</th>\n",
              "      <th>ffa_median</th>\n",
              "      <th>ffa_diff</th>\n",
              "      <th>ggt_median</th>\n",
              "      <th>ggt_diff</th>\n",
              "      <th>glucose_median</th>\n",
              "      <th>glucose_diff</th>\n",
              "      <th>hematocrite_median</th>\n",
              "      <th>hematocrite_diff</th>\n",
              "      <th>hemoglobin_diff</th>\n",
              "      <th>...</th>\n",
              "      <th>pcr_median</th>\n",
              "      <th>pcr_diff</th>\n",
              "      <th>ph_arterial_median</th>\n",
              "      <th>ph_arterial_diff</th>\n",
              "      <th>ph_venous_median</th>\n",
              "      <th>ph_venous_diff</th>\n",
              "      <th>platelets_median</th>\n",
              "      <th>platelets_diff</th>\n",
              "      <th>potassium_median</th>\n",
              "      <th>potassium_diff</th>\n",
              "      <th>sat02_arterial_median</th>\n",
              "      <th>sat02_arterial_diff</th>\n",
              "      <th>sat02_venous_median</th>\n",
              "      <th>sat02_venous_diff</th>\n",
              "      <th>sodium_median</th>\n",
              "      <th>sodium_diff</th>\n",
              "      <th>tgo_median</th>\n",
              "      <th>tgo_diff</th>\n",
              "      <th>tgp_median</th>\n",
              "      <th>tgp_diff</th>\n",
              "      <th>ttpa_median</th>\n",
              "      <th>ttpa_diff</th>\n",
              "      <th>urea_median</th>\n",
              "      <th>urea_diff</th>\n",
              "      <th>dimer_median</th>\n",
              "      <th>dimer_diff</th>\n",
              "      <th>bloodpressure_diastolic_mean</th>\n",
              "      <th>bloodpressure_sistolic_mean</th>\n",
              "      <th>heart_rate_mean</th>\n",
              "      <th>respiratory_rate_mean</th>\n",
              "      <th>temperature_mean</th>\n",
              "      <th>oxygen_saturation_mean</th>\n",
              "      <th>bloodpressure_diastolic_diff</th>\n",
              "      <th>bloodpressure_sistolic_diff</th>\n",
              "      <th>heart_rate_diff</th>\n",
              "      <th>respiratory_rate_diff</th>\n",
              "      <th>temperature_diff</th>\n",
              "      <th>oxygen_saturation_diff</th>\n",
              "      <th>window</th>\n",
              "      <th>icu</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>5</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.183673</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.868365</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.945093</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.090147</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.875236</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.540721</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.518519</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.028571</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.997201</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.990854</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.825613</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.836145</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.994912</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.086420</td>\n",
              "      <td>-0.230769</td>\n",
              "      <td>-0.283019</td>\n",
              "      <td>-0.593220</td>\n",
              "      <td>-0.285714</td>\n",
              "      <td>0.736842</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.912243</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.958528</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.780261</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.144654</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.939887</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.399199</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.703704</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.085714</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.995428</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.986662</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.836145</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.978029</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.489712</td>\n",
              "      <td>-0.685470</td>\n",
              "      <td>-0.048218</td>\n",
              "      <td>-0.645951</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>0.935673</td>\n",
              "      <td>-0.547826</td>\n",
              "      <td>-0.533742</td>\n",
              "      <td>-0.603053</td>\n",
              "      <td>-0.764706</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.959596</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>0</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>-0.263158</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.972789</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.326531</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.968861</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.194030</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.316589</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.203354</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.503592</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.564753</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.777778</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.580247</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.200000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.989549</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.956555</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.937349</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.978029</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.012346</td>\n",
              "      <td>-0.369231</td>\n",
              "      <td>-0.528302</td>\n",
              "      <td>-0.457627</td>\n",
              "      <td>-0.285714</td>\n",
              "      <td>0.684211</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.935113</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.913659</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.829424</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938084</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.851024</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.358491</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.990926</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.457944</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.592593</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.142857</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.998507</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.991235</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.903614</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.333333</td>\n",
              "      <td>-0.153846</td>\n",
              "      <td>0.160377</td>\n",
              "      <td>-0.593220</td>\n",
              "      <td>0.285714</td>\n",
              "      <td>0.868421</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.877301</td>\n",
              "      <td>-0.923664</td>\n",
              "      <td>-0.882353</td>\n",
              "      <td>-0.952381</td>\n",
              "      <td>-0.979798</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>5</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891012</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.958528</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.291405</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.997732</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.292390</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.666667</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.085714</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.997947</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.988948</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.884337</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.037037</td>\n",
              "      <td>-0.538462</td>\n",
              "      <td>-0.537736</td>\n",
              "      <td>-0.525424</td>\n",
              "      <td>-0.196429</td>\n",
              "      <td>0.815789</td>\n",
              "      <td>-0.826087</td>\n",
              "      <td>-0.754601</td>\n",
              "      <td>-0.984733</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.976190</td>\n",
              "      <td>-0.979798</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 97 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "   patient_visit_identifier  age_above65  ...  window  icu\n",
              "0                         0            1  ...     0-2    1\n",
              "1                         2            0  ...     0-2    1\n",
              "2                         3            0  ...     0-2    0\n",
              "3                         4            0  ...     0-2    0\n",
              "4                         5            0  ...     0-2    0\n",
              "\n",
              "[5 rows x 97 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OKmY18RenuHR",
        "outputId": "bb8698c5-183b-4b51-d47b-ecd37363e9cb"
      },
      "source": [
        "# verifica as dimensões do dataframe\n",
        "dimensao_dados(dados)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "O dataframe possui 352 linhas e 97 colunas.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sgrx4nppoFMC"
      },
      "source": [
        "O dataframe utilizado aqui informações de **352 pacientes** e **97 features** (colunas) para alimentar os modelos de Machine Learning.\n",
        "\n",
        "A coluna **patient_visit_identifier** será removida dos dados pois não possui informação relevante para a previsão dos modelos."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "id": "tij4FYXz-qGs",
        "outputId": "f57c029a-d58f-49c5-85a4-43b8cf573e49"
      },
      "source": [
        "# remove a coluna\n",
        "dados_ml = dados.drop('patient_visit_identifier', axis=1)\n",
        "dados_ml.head()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>age_above65</th>\n",
              "      <th>age_percentil</th>\n",
              "      <th>gender</th>\n",
              "      <th>disease grouping 1</th>\n",
              "      <th>disease grouping 2</th>\n",
              "      <th>disease grouping 3</th>\n",
              "      <th>disease grouping 4</th>\n",
              "      <th>disease grouping 5</th>\n",
              "      <th>disease grouping 6</th>\n",
              "      <th>htn</th>\n",
              "      <th>immunocompromised</th>\n",
              "      <th>other</th>\n",
              "      <th>albumin_median</th>\n",
              "      <th>albumin_diff</th>\n",
              "      <th>be_arterial_median</th>\n",
              "      <th>be_arterial_diff</th>\n",
              "      <th>be_venous_median</th>\n",
              "      <th>be_venous_diff</th>\n",
              "      <th>bic_arterial_median</th>\n",
              "      <th>bic_arterial_diff</th>\n",
              "      <th>bic_venous_median</th>\n",
              "      <th>bic_venous_diff</th>\n",
              "      <th>billirubin_median</th>\n",
              "      <th>billirubin_diff</th>\n",
              "      <th>blast_median</th>\n",
              "      <th>blast_diff</th>\n",
              "      <th>calcium_median</th>\n",
              "      <th>calcium_diff</th>\n",
              "      <th>creatinin_median</th>\n",
              "      <th>creatinin_diff</th>\n",
              "      <th>ffa_median</th>\n",
              "      <th>ffa_diff</th>\n",
              "      <th>ggt_median</th>\n",
              "      <th>ggt_diff</th>\n",
              "      <th>glucose_median</th>\n",
              "      <th>glucose_diff</th>\n",
              "      <th>hematocrite_median</th>\n",
              "      <th>hematocrite_diff</th>\n",
              "      <th>hemoglobin_diff</th>\n",
              "      <th>inr_median</th>\n",
              "      <th>...</th>\n",
              "      <th>pcr_median</th>\n",
              "      <th>pcr_diff</th>\n",
              "      <th>ph_arterial_median</th>\n",
              "      <th>ph_arterial_diff</th>\n",
              "      <th>ph_venous_median</th>\n",
              "      <th>ph_venous_diff</th>\n",
              "      <th>platelets_median</th>\n",
              "      <th>platelets_diff</th>\n",
              "      <th>potassium_median</th>\n",
              "      <th>potassium_diff</th>\n",
              "      <th>sat02_arterial_median</th>\n",
              "      <th>sat02_arterial_diff</th>\n",
              "      <th>sat02_venous_median</th>\n",
              "      <th>sat02_venous_diff</th>\n",
              "      <th>sodium_median</th>\n",
              "      <th>sodium_diff</th>\n",
              "      <th>tgo_median</th>\n",
              "      <th>tgo_diff</th>\n",
              "      <th>tgp_median</th>\n",
              "      <th>tgp_diff</th>\n",
              "      <th>ttpa_median</th>\n",
              "      <th>ttpa_diff</th>\n",
              "      <th>urea_median</th>\n",
              "      <th>urea_diff</th>\n",
              "      <th>dimer_median</th>\n",
              "      <th>dimer_diff</th>\n",
              "      <th>bloodpressure_diastolic_mean</th>\n",
              "      <th>bloodpressure_sistolic_mean</th>\n",
              "      <th>heart_rate_mean</th>\n",
              "      <th>respiratory_rate_mean</th>\n",
              "      <th>temperature_mean</th>\n",
              "      <th>oxygen_saturation_mean</th>\n",
              "      <th>bloodpressure_diastolic_diff</th>\n",
              "      <th>bloodpressure_sistolic_diff</th>\n",
              "      <th>heart_rate_diff</th>\n",
              "      <th>respiratory_rate_diff</th>\n",
              "      <th>temperature_diff</th>\n",
              "      <th>oxygen_saturation_diff</th>\n",
              "      <th>window</th>\n",
              "      <th>icu</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>5</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.183673</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.868365</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.945093</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.090147</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.932246</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.875236</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.540721</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.518519</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.028571</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.997201</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.990854</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.825613</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.836145</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.994912</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.086420</td>\n",
              "      <td>-0.230769</td>\n",
              "      <td>-0.283019</td>\n",
              "      <td>-0.593220</td>\n",
              "      <td>-0.285714</td>\n",
              "      <td>0.736842</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.912243</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.958528</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.780261</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.144654</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.959849</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.939887</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.399199</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.703704</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.085714</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.995428</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.986662</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.836145</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.978029</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.489712</td>\n",
              "      <td>-0.685470</td>\n",
              "      <td>-0.048218</td>\n",
              "      <td>-0.645951</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>0.935673</td>\n",
              "      <td>-0.547826</td>\n",
              "      <td>-0.533742</td>\n",
              "      <td>-0.603053</td>\n",
              "      <td>-0.764706</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.959596</td>\n",
              "      <td>0-2</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>-0.263158</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.972789</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.326531</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.968861</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.194030</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.316589</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.203354</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.959849</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.503592</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.564753</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.777778</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.580247</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.200000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.989549</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.956555</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.937349</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.978029</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.012346</td>\n",
              "      <td>-0.369231</td>\n",
              "      <td>-0.528302</td>\n",
              "      <td>-0.457627</td>\n",
              "      <td>-0.285714</td>\n",
              "      <td>0.684211</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.935113</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.913659</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.829424</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938084</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.851024</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.358491</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.959849</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.990926</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.457944</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.592593</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.142857</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.998507</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.991235</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.903614</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.333333</td>\n",
              "      <td>-0.153846</td>\n",
              "      <td>0.160377</td>\n",
              "      <td>-0.593220</td>\n",
              "      <td>0.285714</td>\n",
              "      <td>0.868421</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.877301</td>\n",
              "      <td>-0.923664</td>\n",
              "      <td>-0.882353</td>\n",
              "      <td>-0.952381</td>\n",
              "      <td>-0.979798</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.605263</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.317073</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.938950</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.357143</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891012</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.742004</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.958528</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.891993</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.291405</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.959849</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.997732</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.234043</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.363636</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.292390</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.666667</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.939394</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.345679</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>0.085714</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.997947</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.988948</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.846633</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.884337</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-1.0</td>\n",
              "      <td>-0.037037</td>\n",
              "      <td>-0.538462</td>\n",
              "      <td>-0.537736</td>\n",
              "      <td>-0.525424</td>\n",
              "      <td>-0.196429</td>\n",
              "      <td>0.815789</td>\n",
              "      <td>-0.826087</td>\n",
              "      <td>-0.754601</td>\n",
              "      <td>-0.984733</td>\n",
              "      <td>-1.000000</td>\n",
              "      <td>-0.976190</td>\n",
              "      <td>-0.979798</td>\n",
              "      <td>0-2</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 96 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "   age_above65  age_percentil  gender  ...  oxygen_saturation_diff  window  icu\n",
              "0            1              5       0  ...               -1.000000     0-2    1\n",
              "1            0              0       0  ...               -0.959596     0-2    1\n",
              "2            0              3       1  ...               -1.000000     0-2    0\n",
              "3            0              0       0  ...               -0.979798     0-2    0\n",
              "4            0              0       0  ...               -0.979798     0-2    0\n",
              "\n",
              "[5 rows x 96 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-4aDAAE8_CC8",
        "outputId": "c25cb027-a6e8-4255-ac31-bcdab4f0e0e9"
      },
      "source": [
        "# verifica as dimensões do dataframe\n",
        "dimensao_dados(dados_ml)"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "O dataframe possui 352 linhas e 96 colunas.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BkS77hxj6ift"
      },
      "source": [
        "# Teste dos Modelos de Machine Learning\n",
        "\n",
        "## Modelo Random Forest Classifier\n",
        "\n",
        "O modelo de **Machine Learning Random Forest Classifier** é definido pelo SciKit Learn como um meta estimador que utiliza um número de árvores de decisão em variadas amostras do dataframe e utiliza média para melhorar a eficácia das previsões e controlar o overfitting.\n",
        "\n",
        "Os **hiperparâmetros** informados para o modelo são:\n",
        "\n",
        "**n_estimators**: número de árvores de decisão criadas\n",
        "\n",
        "**criterion**: a função utilizada para medir a qualidade da divisão nas árvores\n",
        "\n",
        "**max_depth**: a máxima profundidade das árvores\n",
        "\n",
        "**max_features**: o número máximo de features a ser considerado em uma divisão das árvores de decisão\n",
        "\n",
        "Abaixo serão testados diferentes hiperparâmetros para o modelo.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "31aPECm8oAMv",
        "outputId": "f62a0e76-977a-4662-dc5c-78c2b867e070"
      },
      "source": [
        "# dicionário com os valores dos hiperparâmetros a serem testados\n",
        "hiper_param_rfc = {\n",
        "    'n_estimators' : [50, 100, 200],\n",
        "    'criterion' : ['gini', 'entropy'],\n",
        "    'max_depth' : [None, 3, 5, 10],\n",
        "    'max_features' : [None, 'auto', 0.5]\n",
        "}\n",
        "\n",
        "# instancia o modelo Random Forest Classifier\n",
        "modelo_rfc = RandomForestClassifier(random_state=481516)\n",
        "\n",
        "# testa os hiperparâmetros\n",
        "teste_hiperparametro(modelo_rfc, dados_ml, hiper_param_rfc)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'criterion': 'entropy',\n",
              "  'max_depth': 5,\n",
              "  'max_features': 'auto',\n",
              "  'n_estimators': 100},\n",
              " 0.8019866292728135)"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dAMxhWvBKYfP"
      },
      "source": [
        "Para o modelo Random Forest Classifier os melhores parâmetros escolhidos por **RandomizedSearchCV** foram:\n",
        "\n",
        "**criterion**: entropy\n",
        "\n",
        "**max_depth**: 5\n",
        "\n",
        "**max_features**: auto\n",
        "\n",
        "**n_estimators**: 100\n",
        "\n",
        "E o **AUC** encontrado pela **validação cruzada** realizado pelo RandomizedSearchCV foi **0.80**.\n",
        "\n",
        "Valores maiores foram testados nos parâmetros **max_depth** e **n_estimators** mas o tempo adicional de processamente não era justificado na melhora do resultado.\n",
        "\n",
        "Com os hiperparâmetros definidos é possível continuar com o modelo."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HnPAdK0i_VVQ"
      },
      "source": [
        "# instancia o modelo\n",
        "modelo_rfc = RandomForestClassifier(criterion='entropy', max_depth=5, max_features='auto',\n",
        "                                    n_estimators=100, random_state=481516)"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CEdLnRZcZUjZ"
      },
      "source": [
        "# função para divisão dos dados em treino e teste\n",
        "x_train, x_test, y_train, y_test = divide_dados(dados_ml)"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TZx0ljy_6NI5"
      },
      "source": [
        "Essa divisão dos dados será utilizada para os demais modelos testados."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xYXI2VNlHfDI"
      },
      "source": [
        "# treina o modelo\n",
        "modelo_rfc.fit(x_train, y_train)\n",
        "\n",
        "# realiza a previsão\n",
        "predicao_rfc = modelo_rfc.predict(x_test)\n",
        "\n",
        "# verifica a probabilidade da previsão\n",
        "prob_prev_rfc = modelo_rfc.predict_proba(x_test)"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JPE9JDblIdY_"
      },
      "source": [
        "### Avaliação do Modelo Random Forest Classifier\n",
        "\n",
        "Após o treinamento e a previsão dos dados é possível verificar o quão bem foi essa previsão com base no dados reais. Para verificar isso serão utilizadas as métricas **Classification Report**, **AUC** e **Matriz de Confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XQBod9iTIWfB",
        "outputId": "0ece9e72-7c70-49ab-da01-1e09b93cfaef"
      },
      "source": [
        "# imprime o Classification Report\n",
        "print('Classification Report:')\n",
        "print(classification_report(y_test, predicao_rfc))"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.73      0.84      0.78        38\n",
            "           1       0.78      0.64      0.70        33\n",
            "\n",
            "    accuracy                           0.75        71\n",
            "   macro avg       0.75      0.74      0.74        71\n",
            "weighted avg       0.75      0.75      0.74        71\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e8Z4UjOqJwnh"
      },
      "source": [
        "O **Classification Report** acima demonstra a performance do modelo de Random Forest:\n",
        "\n",
        "Na coluna **Precision**:\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente não precisa de UTI, o modelo acertou 73%.\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente precisa de UTI, o modelo acertou 78%.\n",
        "\n",
        "Na coluna **Recall**:\n",
        "\n",
        "Considerando os pacientes que não foram para a UTI, o modelo acertou a previsão para 84% desses pacientes.\n",
        "\n",
        "Considerando os pacientes que foram para a UTI, o modelo acertou a previsão para 64% desses pacientes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GGOmw2sXJl88",
        "outputId": "2efd07d7-bce5-4860-9cde-3ea58b64f503"
      },
      "source": [
        "# calcula o AUC do modelo\n",
        "auc = roc_auc_score(y_test, prob_prev_rfc[:, 1])\n",
        "\n",
        "# imprime o valor do AUC\n",
        "print(f'O AUC para o modelo Random Forest Classifier é: {auc}')"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "O AUC para o modelo Random Forest Classifier é: 0.7822966507177034\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 648
        },
        "id": "fPTqkIbjO23E",
        "outputId": "721c8a3c-6b2c-4bbf-a86d-255220c687be"
      },
      "source": [
        "# função para plotar a curva ROC\n",
        "plota_curva_roc(modelo_rfc, 'Random Forest', dados_ml, 'darkgreen')"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wWH-E-jabJu2"
      },
      "source": [
        "As informações acima, **AUC e Curva ROC**, demonstram que o modelo tem um AUC de **0.78**. Esse valor isolado não é ruim, mas quanto é acertado pelo modelo quando paciente precisa ir para a UTI e quando paciente não precisa?\n",
        "\n",
        "A **Matriz de Confusão** pode ser usada para ver essa métrica."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 553
        },
        "id": "lohyE-g-Yvyi",
        "outputId": "e74405ef-50d3-45be-b3f2-76826e4e1206"
      },
      "source": [
        "# plota a matriz de confusão\n",
        "plota_matriz_confusa(modelo_rfc, 'Random Forest', dados_ml, 'Greens')"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CW7GFDIpcL1A"
      },
      "source": [
        "A **matriz** acima mostra os **valores reais** e os que foram **previstos** pelo modelo.\n",
        "\n",
        "Na matriz temos as informações:\n",
        "- 33 pacientes precisaram de UTI, o **modelo classificou corretamente 21 desses pacientes - 64%**.\n",
        "\n",
        "- 38 pacientes não precisaram de UTI, o **modelo classificou corretamente 32 desses pacientes - 84%**.\n",
        "\n",
        "**O modelo não prever corretamente que 12 pessoas (36% do total de pessoas que precisaram de UTI) precisariam de UTI é crítico**.\n",
        "\n",
        "Já sobre os dados de pacientes que não precisaram de UTI, o **modelo errou somente 6 pessoas (16% do total de pacientes que não precisaram de UTI)**.\n",
        "\n",
        "O modelo tem uma porcentagem de acerto boa para dizer se o paciente não precisará de UTI, mas os falsos negativos são críticos nessa situação. Ao dizer que um paciente não precisa de UTI, e na realidade ele necessita, a doença pode se agravar antes do paciente ter os cuidados necessários.\n",
        "\n",
        "Para verificar se esses são os valores encontrados com todos os dados ou somente com essa porção será realizada uma **validação cruzada**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8kioS-M2pJOG",
        "outputId": "1664ef21-6014-49dd-895d-4669dad732f2"
      },
      "source": [
        "# realiza validação cruzada do modelo\n",
        "cv_modelo(modelo_rfc, dados_ml, 5, 10)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC médio para os dados de teste: 0.8026639618086988\n",
            "AUC médio para os dados de treino: 0.9784580485772651\n",
            "Intervalo da AUC para os dados de teste: 0.6979367764546488 - 0.9073911471627488\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.8026639618086988, 0.9784580485772651)"
            ]
          },
          "metadata": {},
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M9AWR-E1qAHS"
      },
      "source": [
        "Na informação acima um **AUC médio de 0.80** foi encontrado nos dados de teste, muito próximo ao encontrado anteriormente (0.78).\n",
        "\n",
        "O **intervalo** informado foi calculado através do **desvio padrão** e mostra uma variação do AUC de **0.70 a 0.90**. Essa variação ocorre devido a porção dos dados que são selecionadas para realizar o treino e teste na validação cruzada."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8I1GB-veodKX"
      },
      "source": [
        "### Modelo Logistic Regression\n",
        "\n",
        "A regressão logística é uma técnica estatística que tem como objetivo produzir, a partir de um conjunto de observações, um modelo que permita a predição de valores tomados por uma variável categórica, frequentemente binária, a partir de uma série de variáveis explicativas contínuas e/ou binárias.\n",
        "\n",
        "Os **hiperparâmetros** informado para o modelo de **Logistic Regression** são:\n",
        "\n",
        "- **C**: parâmetro de regularização inversa, precisa ser um número float positivo;\n",
        "\n",
        "- **solver**: algoritmo a ser utilizado na otimização do problema;\n",
        "\n",
        "- **max_iter**: número máximo de iterações tomadas dos solvers.\n",
        "\n",
        "- **penalty**: especifica a norma a ser utilizada na penalização da regularização."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "orgtCcE2cD8U",
        "outputId": "ecb42e93-dc7f-49c3-ecc5-2d3df2b580ab"
      },
      "source": [
        "# dicionário com os hiperparâmetros\n",
        "hiper_param_lr = {\n",
        "    'penalty' : ['l1', 'l2', 'elasticnet', 'none'],\n",
        "    'C' : [0.2, 0.5, 1.0, 2.0, 3.0],\n",
        "    'solver' : ['lbfgs', 'liblinear', 'newton-cg', 'sag', 'saga'],\n",
        "    'max_iter' : [50, 100, 200, 500]\n",
        "}\n",
        "\n",
        "# instância o modelo\n",
        "modelo_lr = LogisticRegression(random_state=481516)\n",
        "\n",
        "# testa os hiperparâmetros\n",
        "teste_hiperparametro(modelo_lr, dados_ml, hiper_param_lr)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'C': 1.0, 'max_iter': 50, 'penalty': 'l2', 'solver': 'newton-cg'},\n",
              " 0.7693819237898184)"
            ]
          },
          "metadata": {},
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7ob0C_357IAd"
      },
      "source": [
        "Os **hiperparâmetros selecionados** como os melhores para o modelo Logistic Regression foram:\n",
        "\n",
        "- penalty: **l2**\n",
        "\n",
        "- C: **1.0**\n",
        "\n",
        "- solver: **newton-cg**\n",
        "\n",
        "- max_iter: **100**\n",
        "\n",
        "E o AUC encontrado pela validação cruzada realizada pelo RandomizedSearchCV foi **0.76**.\n",
        "\n",
        "*Ao tentar executar o modelo com o valor 50 do max_iter, um warning aconteceu informando que o número de iterações atingiu o limite. Por esse motivo o valor foi alterado para 100.*"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QcgNXmlmv6Ja"
      },
      "source": [
        "# instancia o modelo com os hiperparâmetros informados\n",
        "modelo_lr = LogisticRegression(penalty='l2', C=1.0, solver='newton-cg',\n",
        "                               max_iter=100, random_state=481516)"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "E7GlfEdw-KRl"
      },
      "source": [
        "Os dados já foram divididos para o modelo Random Forest, não há necessidade de dividir novamente."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ivmLMKOq-QRI"
      },
      "source": [
        "# treina o modelo\n",
        "modelo_lr.fit(x_train, y_train)\n",
        "\n",
        "# realiza a previsão\n",
        "previsao_lr = modelo_lr.predict(x_test)\n",
        "\n",
        "# verifica a probabilidade da previsão\n",
        "prob_prev_lr = modelo_lr.predict_proba(x_test)"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uDxmFRaO_ccF"
      },
      "source": [
        "### Avaliação do modelo Logistic Regression\n",
        "\n",
        "Como para o modelo Random Forest, aqui serão avaliadas as métricas informadas pelo **Classification Report**, **ROC AUC** e **Matriz de Confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "l85rdKWF-wPE",
        "outputId": "8ed9099d-6c72-422c-b75b-c958eccdbf8a"
      },
      "source": [
        "# imprime o Classification Report\n",
        "print('Classification Report:\\n')\n",
        "print(classification_report(y_test, previsao_lr))"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Classification Report:\n",
            "\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.72      0.82      0.77        38\n",
            "           1       0.75      0.64      0.69        33\n",
            "\n",
            "    accuracy                           0.73        71\n",
            "   macro avg       0.74      0.73      0.73        71\n",
            "weighted avg       0.73      0.73      0.73        71\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uAHm6nfNM14v"
      },
      "source": [
        "O **Classification Report** acima demonstra a performance do modelo de **Logistic Regression**:\n",
        "\n",
        "Na coluna **Precision**:\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente não precisa de UTI, o modelo acertou 72%.\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente precisa de UTI, o modelo acertou 75%.\n",
        "\n",
        "Na coluna **Recall**:\n",
        "\n",
        "Considerando os pacientes que não foram para a UTI, o modelo acertou a previsão para 82% desses pacientes.\n",
        "\n",
        "Considerando os pacientes que foram para a UTI, o modelo acertou a previsão para 64% desses pacientes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2JJfM0IHNAi8",
        "outputId": "3507a1c2-d205-44cd-f7d9-405ae024c933"
      },
      "source": [
        "# calcula o AUC do modelo\n",
        "auc = roc_auc_score(y_test, prob_prev_lr[:, 1])\n",
        "\n",
        "# imprime o valor do AUC\n",
        "print(f'O AUC para o modelo Random Forest Classifier é: {auc}')"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "O AUC para o modelo Random Forest Classifier é: 0.7416267942583733\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 648
        },
        "id": "rK8U7_m1NI2H",
        "outputId": "85e8db0f-bfb1-42d6-c501-14eee81f6ca9"
      },
      "source": [
        "# função para plotar a curva ROC\n",
        "plota_curva_roc(modelo_lr, 'Logistic Regression', dados_ml, 'maroon')"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fQZIK6HENnQY"
      },
      "source": [
        "Para o modelo Logistic Regression o **AUC** encontrado é **0.74**. Abaixo as previsões serão detalhadas na **matriz de confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 553
        },
        "id": "0ER7SJoPNYfZ",
        "outputId": "9c7cb4f0-48f2-409e-ad02-9286c1c2934a"
      },
      "source": [
        "# plota a matriz de confusão\n",
        "plota_matriz_confusa(modelo_lr, 'Logistic Regression', dados_ml, 'Reds')"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1zaEVewAOQ9G"
      },
      "source": [
        "A matriz acima mostra os **valores reais** e os que foram **previstos pelo modelo**.\n",
        "\n",
        "Na matriz temos as informações:\n",
        "- 33 pacientes precisaram de UTI, o **modelo classificou corretamente 21 desses pacientes - 64%**.\n",
        "\n",
        "- 38 pacientes não precisaram de UTI, o **modelo classificou corretamente 31 desses pacientes - 82%**.\n",
        "\n",
        "O modelo não previu corretamente que 12 pessoas (36% do total de pessoas que precisaram de UTI) precisariam de UTI é crítico.\n",
        "\n",
        "Já sobre os dados de pacientes que não precisaram de UTI, o **modelo errou 7 pessoas (18% do total de pacientes que não precisaram de UTI)**.\n",
        "\n",
        "O modelo tem uma porcentagem de acerto boa para dizer se o paciente não precisará de UTI, mas os falsos negativos são críticos nessa situação. Ao dizer que um paciente não precisa de UTI, e na realidade ele necessita, a doença pode se agravar antes do paciente ter os cuidados necessários.\n",
        "\n",
        "Para verificar se esses são os valores encontrados com todos os dados ou somente com essa porção será realizada uma **validação cruzada**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TYGxjNyIOC5p",
        "outputId": "3ad0de84-810b-4397-807f-af9f15640a31"
      },
      "source": [
        "# realiza validação cruzada do modelo\n",
        "cv_modelo(modelo_lr, dados_ml, 5, 10)"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC médio para os dados de teste: 0.7697778729686625\n",
            "AUC médio para os dados de treino: 0.8518187752038671\n",
            "Intervalo da AUC para os dados de teste: 0.675589497140343 - 0.863966248796982\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.7697778729686625, 0.8518187752038671)"
            ]
          },
          "metadata": {},
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sMXZxZZnO53f"
      },
      "source": [
        "Na informação acima um **AUC médio de 0.77** foi encontrado nos dados de teste, muito próximo ao encontrado anteriormente (**0.74**).\n",
        "\n",
        "O **intervalo** informado foi calculado através do **desvio padrão** e mostra uma variação do **AUC de 0.67 a 0.86**. Essa variação ocorre devido a porção dos dados que são selecionadas para realizar o treino e teste na validação cruzada.\n",
        "\n",
        "De modo geral o modelo **Random Forest** teve uma performance **MELHOR** que o **Logistic Regression**."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MAjqIf5IPPuI"
      },
      "source": [
        "# Modelos SVM (Support Vector Machine)\n",
        "\n",
        "## Modelo SVC (Support Vector Classification)\n",
        "\n",
        "Modelos **SVM** são métodos de aprendizado supervisionado utilizados para classificação, regressão e detecção de outliers. Algumas vantagens desses modelos são:\n",
        "\n",
        "- efetivo em espações multidimensionais;\n",
        "\n",
        "- efetivo em casos onde o número de dimensões é maior que o número de amostras;\n",
        "\n",
        "- utilizado uma porção dos dados de treino na função de decisão (chamados vetores de suporte);\n",
        "\n",
        "Os **hiperparâmetros** informados para o modelo são:\n",
        "\n",
        "- **C**: parâmetro de regularização inversa, precisa ser um número float positivo;\n",
        "\n",
        "- **kernel**: informa o tipo de kernel a ser utilizado pelo algoritmo;\n",
        "\n",
        "- **degree**: grau da função polinomial do kernel ('poly').\n",
        "\n",
        "- **gamma**: coeficiente do kernel para 'rbf', 'poly' e 'sigmoid';"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UMhHB7KDO0e6",
        "outputId": "b9779208-90e5-4d7e-a8f3-5a692fa37afa"
      },
      "source": [
        "# dicionário com os hiperparâmetros\n",
        "hiper_param_svc = {\n",
        "    'C' : [0.5, 1.0, 3.0, 5.0, 10.0],\n",
        "    'kernel' : ['linear', 'poly', 'rbf', 'sigmoid'],\n",
        "    'degree' : [1, 3, 5, 7],\n",
        "    'gamma' : ['scale', 'auto']\n",
        "}\n",
        "\n",
        "# instancia o modelo\n",
        "modelo_svc = SVC(random_state=481516)\n",
        "\n",
        "# função para testar os hiperparâmetros\n",
        "teste_hiperparametro(modelo_svc, dados_ml, hiper_param_svc)"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'C': 3.0, 'degree': 3, 'gamma': 'scale', 'kernel': 'poly'},\n",
              " 0.7713704523901893)"
            ]
          },
          "metadata": {},
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BlWTOz8DS852"
      },
      "source": [
        "Os **hiperparâmetros** selecionados pelo **RandomizedSearchCV** são:\n",
        "\n",
        "- C: **3.0**;\n",
        "\n",
        "- kernel: **poly**;\n",
        "\n",
        "- degree: **3**;\n",
        "\n",
        "- gamma: **scale**;\n",
        "\n",
        "O AUC encontrado pela validação cruzada realizada pelo RandomizedSearchCV foi **0.77**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RCIz3UzbSk3D"
      },
      "source": [
        "# instancia o modelo com os hiperparâmetros selecionados\n",
        "modelo_svc = SVC(C=3.0, kernel='poly', degree=3, gamma='scale', random_state=481516,\n",
        "                 probability=True)"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7FE-TqGjT_h4"
      },
      "source": [
        "Os dados já foram divididos anteriormente.\n",
        "\n",
        "O parâmetro **probability** foi adicionado para que o **predict_proba** seja informado."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sjypVkrKT-sv"
      },
      "source": [
        "# treina o modelo\n",
        "modelo_svc.fit(x_train, y_train)\n",
        "\n",
        "# realiza a previsão\n",
        "previsao_svc = modelo_svc.predict(x_test)\n",
        "\n",
        "# verifica a probabilidade da previsão\n",
        "prob_prev_svc = modelo_svc.predict_proba(x_test)"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "spG2UKhGUdih"
      },
      "source": [
        "### Avaliação do modelo SVC\n",
        "\n",
        "Como para os modelos anteriores, aqui serão avaliadas as métricas informadas pelo **Classification Report**, **ROC AUC** e **Matriz de Confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "U-U8rSv-UHqG",
        "outputId": "7cde80e1-470d-4802-8984-47b9fcb3a4a7"
      },
      "source": [
        "# imprime o Classification Report\n",
        "print('Classification Report:\\n')\n",
        "print(classification_report(y_test, previsao_svc))"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Classification Report:\n",
            "\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.69      0.76      0.72        38\n",
            "           1       0.69      0.61      0.65        33\n",
            "\n",
            "    accuracy                           0.69        71\n",
            "   macro avg       0.69      0.68      0.69        71\n",
            "weighted avg       0.69      0.69      0.69        71\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V1NC4qbdVBUn"
      },
      "source": [
        "O **Classification Report** acima demonstra a performance do modelo de **SVC**:\n",
        "\n",
        "Na coluna **Precision**:\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente não precisa de UTI, o modelo acertou 69%.\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente precisa de UTI, o modelo acertou 69%.\n",
        "\n",
        "Na coluna **Recall**:\n",
        "\n",
        "Considerando os pacientes que não foram para a UTI, o modelo acertou a previsão para 76% desses pacientes.\n",
        "\n",
        "Considerando os pacientes que foram para a UTI, o modelo acertou a previsão para 61% desses pacientes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 648
        },
        "id": "R_YZYa-8Uq68",
        "outputId": "b82999a1-31b3-47ed-c0bc-cb8cf16dad5f"
      },
      "source": [
        "# função para plotar a curva ROC\n",
        "plota_curva_roc(modelo_svc, 'SVC', dados_ml, 'cornflowerblue')"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Yn8zfYrRVhI2"
      },
      "source": [
        "Para o modelo SVC o **AUC** encontrado é **0.75**. Abaixo as previsões serão detalhadas na **matriz de confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 553
        },
        "id": "jgNTm45jVZZ4",
        "outputId": "74df1c3e-ad29-4381-f3e2-26c352524c88"
      },
      "source": [
        "# função para plotar a matriz de confusão\n",
        "plota_matriz_confusa(modelo_svc, 'SVC', dados_ml, 'Blues')"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BK6EXGEnV-N3"
      },
      "source": [
        "A **matriz** acima mostra os **valores reais** e os que **foram previstos pelo modelo**.\n",
        "\n",
        "Na matriz temos as informações:\n",
        "- 33 pacientes precisaram de UTI, o **modelo classificou corretamente 20 desses pacientes - 61%**.\n",
        "\n",
        "- 38 pacientes não precisaram de UTI, o **modelo classificou corretamente 29 desses pacientes - 76%**.\n",
        "\n",
        "O modelo não previu corretamente que 13 pessoas (39% do total de pessoas que precisaram de UTI) precisariam de UTI é crítico.\n",
        "\n",
        "Já sobre os dados de pacientes que não precisaram de UTI, o **modelo errou 9 pessoas (24% do total de pacientes que não precisaram de UTI)**.\n",
        "\n",
        "No geral o modelo **NÃO** teve uma performance boa para prever se o paciente precisará ou não de UTI.\n",
        "\n",
        "Para verificar se esses são os valores encontrados com todos os dados ou somente com essa porção será realizada uma **validação cruzada**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Khs5WXlSV4gP",
        "outputId": "f3bd1ae4-8819-41c9-92ca-70bb642ba640"
      },
      "source": [
        "# realiza validação cruzada do modelo\n",
        "cv_modelo(modelo_svc, dados_ml, 5, 10)"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC médio para os dados de teste: 0.7715038229018493\n",
            "AUC médio para os dados de treino: 0.8645495207952179\n",
            "Intervalo da AUC para os dados de teste: 0.6837006693069645 - 0.859306976496734\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.7715038229018493, 0.8645495207952179)"
            ]
          },
          "metadata": {},
          "execution_count": 35
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CVLgd1tEWsv2"
      },
      "source": [
        "A validação cruzada encontrou um **AUC médio de 0.77** nos dados de teste, muito próximo ao encontrado anteriormente (**0.75**).\n",
        "\n",
        "O **intervalo** informado foi calculado através do **desvio padrão** e mostra uma variação do **AUC de 0.68 a 0.85**. Essa variação ocorre devido a porção dos dados que são selecionadas para realizar o treino e teste na validação cruzada.\n",
        "\n",
        "De modo geral o modelo SVC teve uma performance **PIOR** que os modelos anteriores."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Mv8Ii6DdXL3m"
      },
      "source": [
        "## Modelo LinearSVC (Linear Supporte Vector Classification)\n",
        "\n",
        "Similar ao modelo SVC porém com o parâmetro kernel definido como linear.\n",
        "\n",
        "Os **hiperparâmetros** informados para o modelo são:\n",
        "\n",
        "- **penalty**: especifica a norma usada na penalização;\n",
        "\n",
        "- **dual**: seleciona o algoritmo para resolver o problema de otimização original (primal) ou duplo (dual);\n",
        "\n",
        "- **C**: parâmetro de regularização;\n",
        "\n",
        "- **max_iter**: número máximo de iteração que serão processadas;"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kUG6EPw0Wf84",
        "outputId": "3c0f0f32-8b69-4cce-8f6c-2b6faa4e20a7"
      },
      "source": [
        "# dicionário com os hiperparâmetros\n",
        "hiper_param_linear_svc = {\n",
        "    'penalty' : ['l1', 'l2'],\n",
        "    'dual' : [True, False],\n",
        "    'C' : [0.1, 0.3, 0.5],\n",
        "    'max_iter' : [300, 500, 1000]\n",
        "}\n",
        "\n",
        "# instancia o modelo Linear SVC\n",
        "modelo_linear_svc = LinearSVC(random_state=481516)\n",
        "\n",
        "# testa os hiperparâmetros\n",
        "teste_hiperparametro(modelo_linear_svc, dados_ml, hiper_param_linear_svc)"
      ],
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "({'C': 0.1, 'dual': False, 'max_iter': 500, 'penalty': 'l2'},\n",
              " 0.7689890323936378)"
            ]
          },
          "metadata": {},
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nTPJRs6iZnDM"
      },
      "source": [
        "Os **hiperparâmetros** selecionados pelo **RandomizedSearchCV** são:\n",
        "\n",
        "- penalty: **l2**;\n",
        "\n",
        "- dual: **False**;\n",
        "\n",
        "- C: **0.1**;\n",
        "\n",
        "- max_iter: **500**;\n",
        "\n",
        "O AUC encontrado pela validação cruzada realizada pelo RandomizedSearchCV foi **0.77**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0DKIT-ivZKwP"
      },
      "source": [
        "# instancia o modelo com os hiperparâmetros selecionados\n",
        "modelo_linear_svc = LinearSVC(penalty='l2', dual=False, C=0.1, max_iter=500,\n",
        "                              random_state=481516)"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4OfEyf_fascZ"
      },
      "source": [
        "Os dados já foram separados em treino e teste anteriormente."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qtB1_wmSal-m"
      },
      "source": [
        "# treina o modelo\n",
        "modelo_linear_svc.fit(x_train, y_train)\n",
        "\n",
        "# realiza a previsão\n",
        "previsao_linear_svc = modelo_linear_svc.predict(x_test)"
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wIvuqbxabOq6"
      },
      "source": [
        "O modelo **LinearSVC** não realiza o cálculo do **predict_proba**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "14tCSSf8a22t",
        "outputId": "6c09c792-4a5d-4078-cf29-79ef6c496e9e"
      },
      "source": [
        "# imprime o Classification Report\n",
        "print('Classification Report:\\n')\n",
        "print(classification_report(y_test, previsao_linear_svc))"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Classification Report:\n",
            "\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.72      0.82      0.77        38\n",
            "           1       0.75      0.64      0.69        33\n",
            "\n",
            "    accuracy                           0.73        71\n",
            "   macro avg       0.74      0.73      0.73        71\n",
            "weighted avg       0.73      0.73      0.73        71\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vSeBffA5dMRe"
      },
      "source": [
        "O **Classification Report** acima demonstra a performance do modelo de **LinearSVC**:\n",
        "\n",
        "Na coluna **Precision**:\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente não precisa de UTI, o modelo acertou 72%.\n",
        "\n",
        "Dentro de todas as previsões indicando que o paciente precisa de UTI, o modelo acertou 75%.\n",
        "\n",
        "Na coluna **Recall**:\n",
        "\n",
        "Considerando os pacientes que não foram para a UTI, o modelo acertou a previsão para 82% desses pacientes.\n",
        "\n",
        "Considerando os pacientes que foram para a UTI, o modelo acertou a previsão para 64% desses pacientes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 648
        },
        "id": "nzlfSbUJdE6j",
        "outputId": "b76edfa1-b074-4ea6-edbb-4580f7ab9afd"
      },
      "source": [
        "# função para plotar a curva ROC\n",
        "plota_curva_roc(modelo_linear_svc, 'LinearSVC', dados_ml, 'darkgoldenrod')"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6E1L0bJAd3tE"
      },
      "source": [
        "Para o modelo LinearSVC o **AUC** encontrado é **0.74**. Abaixo as previsões serão detalhadas na **matriz de confusão**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 553
        },
        "id": "z00uHn89dtNS",
        "outputId": "5943b05d-8d94-4fff-beb8-0fc139985c1d"
      },
      "source": [
        "# função para plotar a matriz de confusão\n",
        "plota_matriz_confusa(modelo_linear_svc, 'LinearSVC', dados_ml, 'YlOrBr')"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B_UE0hOCeL0k"
      },
      "source": [
        "A **matriz** acima mostra os **valores reais** e os que **foram previstos pelo modelo**.\n",
        "\n",
        "Na matriz temos as informações:\n",
        "- 33 pacientes precisaram de UTI, o **modelo classificou corretamente 21 desses pacientes - 64%**.\n",
        "\n",
        "- 38 pacientes não precisaram de UTI, o **modelo classificou corretamente 31 desses pacientes - 82%**.\n",
        "\n",
        "O modelo não previu corretamente que 12 pessoas (36% do total de pessoas que precisaram de UTI) precisariam de UTI é crítico.\n",
        "\n",
        "Já sobre os dados de pacientes que não precisaram de UTI, o **modelo errou 7 pessoas (18% do total de pacientes que não precisaram de UTI)**.\n",
        "\n",
        "O modelo teve uma boa performance em prever os pacientes que não precisaram de UTI, porém o erro de previsão para 12 pacientes que precisaram de UTI é um ponto a ser considerado.\n",
        "\n",
        "Para verificar se esses são os valores encontrados com todos os dados ou somente com essa porção será realizada uma **validação cruzada**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yQ9DUjLTeGzp",
        "outputId": "bda177d6-03c5-4eb5-d79d-b843c5f74dc6"
      },
      "source": [
        "# realiza validação cruzada do modelo\n",
        "cv_modelo(modelo_linear_svc, dados_ml, 5, 10)"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC médio para os dados de teste: 0.7700455569205569\n",
            "AUC médio para os dados de treino: 0.8506038276004844\n",
            "Intervalo da AUC para os dados de teste: 0.6767899075898329 - 0.8633012062512808\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.7700455569205569, 0.8506038276004844)"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jrLOKErdfgqh"
      },
      "source": [
        "A **validação cruzada** encontrou um **AUC médio de 0.77** nos dados de teste, muito próximo ao encontrado anteriormente (**0.74**).\n",
        "\n",
        "O **intervalo** informado foi calculado através do **desvio padrão** e mostra uma variação do **AUC de 0.68 a 0.86**. Essa variação ocorre devido a porção dos dados que são selecionadas para realizar o treino e teste na validação cruzada.\n",
        "\n",
        "De modo geral o modelo LinearSVC teve uma performance **MELHOR** que o modelo SVC, com resultados muito próximos aos encontrados com os modelos Random Forest e Logistic Regression."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yfbfRcm9gCZR"
      },
      "source": [
        "# Conclusão\n",
        "\n",
        "Chegando ao fim desse notebook, e por consequência ao final desse projeto, as conclusões abaixo foram constatadas:\n",
        "\n",
        "- 4 modelos de classificação foram testados com os dados disponibilizados pelo Hospital Sírio Libanês com o objetivo de prever a necessidade, ou não, de um paciente necessitar de UTI;\n",
        "\n",
        "- Entre os 4 modelos testados, o melhor desempenho foi apresentado pelo modelo **Random Forest Classifier**, seguido do **LinearSVC** e **Logistic Regression** que apresentaram resultados similares. Abaixo os modelos estão ordenados por desempenho:\n",
        "\n",
        "1. **Random Forest Classifier**\n",
        "\n",
        "2. **LinearSVC**\n",
        "\n",
        "3. **Logistic Regression**\n",
        "\n",
        "4. **SVC**\n",
        "\n",
        "### Comparação das métricas de desempenho entre os modelos\n",
        "\n",
        "| Métrica de Desempenho |Random Forest Classifier|LinearSVC|Logistic Regression|SVC|\n",
        "|:---------------------:|:----------------------:|:-------:|:-----------:|:-----:|\n",
        "|PRECISION(0)           | **0.73** | **0.72** | **0.72** | **0.69** |\n",
        "|PRECISION(1)           | **0.78** | **0.75** | **0.75** | **0.69** |\n",
        "|RECALL(0)              | **0.84** | **0.82** | **0.82** | **0.76** |\n",
        "|RECALL(1)              | **0.64** | **0.64** | **0.64** | **0.61** |\n",
        "|F1.Score(0)            | **0.78** | **0.77** | **0.77** | **0.72** |\n",
        "|F1.Score(1)            | **0.70** | **0.69** | **0.69** | **0.65** |\n",
        "|AUC TEST               | **0.80** | **0.77** | **0.77** | **0.77** |\n",
        "|AUC TREINO             | **0.98** | **0.85** | **0.85** | **0.86** |\n",
        "|AUC                    | **0.78** | **0.74** | **0.74** | **0.75** |\n",
        "\n",
        "O desempenho encontrado por alguns dos modelos testados nesse projeto são satisfatórios e podem auxiliar no diagnóstico da necessidade de um leito de UTI por um paciente assim que chegar ao hospital.\n",
        "\n",
        "Como os modelos Random Forest Classifiers, LinearSVC e Logistic Regression apresentaram resultados próximos, uma solução interessante seria criar um **ensemble** com os 3 modelos e verificar se os resultados encontrados seriam melhores."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3m4B7IOhmAEO"
      },
      "source": [
        "## Próximos Passos\n",
        "\n",
        "As seguintes sugestões podem ser utilizadas para melhorar o desempenho das previsões dos modelos testados nesse projeto:\n",
        "\n",
        "- Testar outros modelos de classificação e verificar as previsões;\n",
        "\n",
        "- Criar um *ensemble* dos modelos com a melhor performance que foram testados nesse projeto;\n",
        "\n",
        "- Testar os modelos adicionando as outras janelas (coluna window) disponíveis nos dados;"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3Wq3aHIdoKE9"
      },
      "source": [
        "## Referências utilizadas durante o projeto\n",
        "\n",
        "1. [Página do Kaggle com informações sobre a base de dados](https://www.kaggle.com/S%C3%ADrio-Libanes/covid19)\n",
        "\n",
        "2. [Text classification with extremely small datasets](https://towardsdatascience.com/text-classification-with-extremely-small-datasets-333d322caee2)\n",
        "\n",
        "3. [5 Feature Selection Method From SciKit Learn You Should Know](https://towardsdatascience.com/5-feature-selection-method-from-scikit-learn-you-should-know-ed4d116e4172)\n",
        "\n",
        "4. [Como Selecionar As Melhores Features Para Seu Modelo de Machine Learning](https://paulovasconcellos.com.br/como-selecionar-as-melhores-features-para-seu-modelo-de-machine-learning-2e9df83d062a)\n",
        "\n",
        "5. [Machine Learning Map](https://scikit-learn.org/stable/tutorial/machine_learning_map/)\n",
        "\n",
        "6. [Primeiro caso de Covid-19 pode ter atingindo a Chine em outubro](https://www.cnnbrasil.com.br/saude/primeiro-caso-covid-19-pode-ter-atingido-a-china-em-outubro-de-2019-diz-estudo/)\n",
        "\n",
        "7. [Evolução dos Sintomas do Covid-19](https://saude.abril.com.br/medicina/casos-sem-sintomas-evolucoes-coronavirus/)\n",
        "\n",
        "8. [Covid Impact to Global Healthcare](https://edition.cnn.com/videos/world/2021/04/14/covid-impact-global-healthcare-lon-orig.cnn)\n",
        "\n",
        "9. [Observatório Covid-19 aponta maior colapso sanitário e hospitalar da história do Brasil](https://portal.fiocruz.br/noticia/observatorio-covid-19-aponta-maior-colapso-sanitario-e-hospitalar-da-historia-do-brasil)\n",
        "\n",
        "10. [Primeiro caso Covid-19 no Brasil](https://saude.abril.com.br/medicina/coronavirus-primeiro-caso-brasil/)\n",
        "\n",
        "11. [Overfitting e Underfitting em Machine Learning](https://abracd.org/overfitting-e-underfitting-em-machine-learning/#:~:text=Underfitting%20%C3%A9%20uma%20tradu%C3%A7%C3%A3o%20para,Overfitting%20%C3%A9%20o%20oposto)\n",
        "\n",
        "12. [WHO delivers advice and support for older people during Covid-19](https://www.who.int/news-room/feature-stories/detail/who-delivers-advice-and-support-for-older-people-during-covid-19#:~:text=The%20COVID%2D19%20pandemic,potential%20underlying%20health%20conditions)\n",
        "\n",
        "13. [Machine Learning in Healthcare](https://healthinformatics.uic.edu/blog/machine-learning-in-healthcare/)\n",
        "\n",
        "14. [Machine Learning](https://www.ibm.com/cloud/learn/machine-learning)\n",
        "\n",
        "15. [Hiperparameter Tuning](https://towardsdatascience.com/hyperparameter-tuning-c5619e7e6624)\n",
        "\n",
        "16. [SKLearn Model Selection - RandomizedSearchCV](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.RandomizedSearchCV.html)\n",
        "\n",
        "17. [Entenda o que é AUC e ROC nos modelos de Machine Learning](https://medium.com/bio-data-blog/entenda-o-que-%C3%A9-auc-e-roc-nos-modelos-de-machine-learning-8191fb4df772)\n",
        "\n",
        "18. [Understanding Confusion Matrix](https://towardsdatascience.com/understanding-confusion-matrix-a9ad42dcfd62)\n",
        "\n",
        "19. [Understanding a Classification Report For Your Machine Learning Model](https://medium.com/@kohlishivam5522/understanding-a-classification-report-for-your-machine-learning-model-88815e2ce397)\n",
        "\n",
        "20. [Understanding AUC ROC Curve](https://towardsdatascience.com/understanding-auc-roc-curve-68b2303cc9c5)\n",
        "\n",
        "21. [Regressão Logística](https://pt.wikipedia.org/wiki/Regress%C3%A3o_log%C3%ADstica)\n",
        "\n",
        "22. [SciKit Learn - RandomForestClassifier](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html)\n",
        "\n",
        "23. [SciKit Learn - Logistic Regression](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html)\n",
        "\n",
        "24. [SciKit Learn - SVC](http://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html)\n",
        "\n",
        "25. [SciKit Learn - LinearSVC](https://scikit-learn.org/stable/modules/generated/sklearn.svm.LinearSVC.html)"
      ]
    }
  ]
}