3.2-GridSearch_KNN_Iris.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyOqbHn38Q1tje1kZOIqXkbD"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "#Grid Search para otimização de hiperparâmetros\n",
        "\n",
        "O GridSearch é uma técnica utilizada para ajustar os parâmetros de um modelo de aprendizado de máquina. Ele funciona criando uma grade de todas as combinações possíveis dos valores dos parâmetros que você deseja otimizar, e treina um modelo para cada combinação de parâmetros. Ele então avalia cada modelo usando uma métrica de desempenho que você especifica e seleciona o modelo com o melhor desempenho.\n",
        "\n",
        "Para usar o GridSearch no scikit-learn, você primeiro precisa definir os parâmetros que deseja otimizar e os valores que deseja testar para cada parâmetro. Em seguida, crie um objeto GridSearchCV passando os valores dos parâmetros e o modelo que você deseja otimizar. Por fim, chame o método fit do objeto GridSearchCV passando os dados de treinamento, e ele treinará um modelo para cada combinação de parâmetros e selecionará o melhor modelo baseado na métrica de desempenho especificada."
      ],
      "metadata": {
        "id": "z3BcUEsCJ1IZ"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LrYqdQQunE0i"
      },
      "outputs": [],
      "source": [
        "from sklearn import datasets\n",
        "from sklearn.model_selection import cross_val_score\n",
        "from sklearn.model_selection import StratifiedKFold\n",
        "from sklearn.model_selection import GridSearchCV\n",
        "\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "\n",
        "import time"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "iris = datasets.load_iris()\n",
        "X = iris.data\n",
        "y = iris.target"
      ],
      "metadata": {
        "id": "VqVCuX1x0sPA"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Para configurar os parâmetros que o KNN irá utilizar para fazer a busca, defini a quantidade de visinhos como **3**, **5** e **7**, os pesos como **Uniforme** e **Distância** (https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html) e a métrica de distância como **Euclideana** e **Manhattan** (https://scikit-learn.org/stable/modules/generated/sklearn.metrics.pairwise.distance_metrics.html#sklearn.metrics.pairwise.distance_metrics)\n",
        "\n",
        "$3 visinhos * 2 pesos * 2 métricas$ totaliza uma busca por 12 combinações diferentes de parâmetros possíveis para essa configuração. Se aumentar a busca de vizinhos para 4 opções, o total salta de 12 para 16 combinações ($4 vizinhos * 2 pesos * 2 métricas = 16 combinações$).\n",
        "\n",
        "Tudo isso, multiplicado pela quantidade de pastas que foram definidas no *Cross Validation*. Neste caso são 10 pastas, então o total de combinações possíveis é multiplicado por 10, chegando a 120 iterações."
      ],
      "metadata": {
        "id": "gd-EJp3u3HOy"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "param_grid = [{'n_neighbors':[3,5,7],\n",
        "               'weights':['uniform', 'distance'],\n",
        "               'metric':['euclidean','manhattan']}]"
      ],
      "metadata": {
        "id": "oCqI98BF0sRm"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "knn = KNeighborsClassifier()\n",
        "\n",
        "inicio = time.time()\n",
        "\n",
        "grid_search_knn = GridSearchCV(knn, param_grid, cv=10, verbose=3 )\n",
        "resultado_modelo = grid_search_knn.fit(X,y)\n",
        "\n",
        "termino = time.time()\n",
        "print(\"--- %s segundos para treinar o modelo ---\" % (termino - inicio))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "81qkuFCX0sUE",
        "outputId": "7092c198-340c-4fcb-e97e-75a6e4f48dca"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Fitting 10 folds for each of 12 candidates, totalling 120 fits\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=euclidean, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=3, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=3, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=5, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=5, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=0.867 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=0.933 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=7, weights=uniform;, score=1.000 total time=   0.0s\n",
            "[CV 1/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 2/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 3/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 4/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 5/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=0.867 total time=   0.0s\n",
            "[CV 6/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 7/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=0.933 total time=   0.0s\n",
            "[CV 8/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 9/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "[CV 10/10] END metric=manhattan, n_neighbors=7, weights=distance;, score=1.000 total time=   0.0s\n",
            "--- 0.4910712242126465 segundos para treinar o modelo ---\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"Melhores Parâmetros\")\n",
        "print(grid_search_knn.best_params_)\n",
        "print(\"**************\")\n",
        "print(\"Melhores Estimadores\")\n",
        "print(grid_search_knn.best_estimator_)\n",
        "print(\"**************\")\n",
        "print(\"Melhores Pontuações\")\n",
        "print(grid_search_knn.best_score_)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Njbjkt1L1DBX",
        "outputId": "371ee7fa-bc80-46f6-aee1-c62295b5fd7f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Melhores Parâmetros\n",
            "{'metric': 'euclidean', 'n_neighbors': 5, 'weights': 'uniform'}\n",
            "**************\n",
            "Melhores Estimadores\n",
            "KNeighborsClassifier(metric='euclidean')\n",
            "**************\n",
            "Melhores Pontuações\n",
            "0.9666666666666668\n"
          ]
        }
      ]
    }
  ]
}