Untitled0.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled0.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyN21ZHckScrMFtHDNumZ35/",
      "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/alessio76/Progetto-ML/blob/main/Untitled0.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0LW9CytfsYxL"
      },
      "source": [
        "# Moduli di scikit-learn\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.model_selection import cross_val_score, GridSearchCV, ParameterGrid, train_test_split\n",
        "from sklearn.datasets import load_breast_cancer\n",
        "from sklearn.neighbors import DistanceMetric\n",
        "from sklearn import metrics\n",
        "from sklearn.preprocessing import StandardScaler, MinMaxScaler, MaxAbsScaler\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.decomposition import PCA\n",
        "import numpy as np\n",
        "from numpy import linalg as la\n",
        "from sklearn import metrics\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns; sns.set()\n",
        "\n"
      ],
      "execution_count": 99,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xkCbtunaurl5"
      },
      "source": [
        "#dataset è un dizionario\n",
        "dataset = load_breast_cancer()\n",
        "x,y = dataset[\"data\"], dataset[\"target\"]\n",
        "train_x, test_x, train_y, test_y = train_test_split(x, y, test_size=.4, random_state=1)"
      ],
      "execution_count": 87,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 86
        },
        "id": "UkrI8KztyvyQ",
        "outputId": "4d504c65-eb20-4466-a4c4-90f3668c014f"
      },
      "source": [
        "classifier =  KNeighborsClassifier()\n",
        "n_folds = 9\n",
        "param_grid = [{'n_neighbors' : [1,3,5,7,9], 'metric' : ['euclidean']},\n",
        "              {'n_neighbors' : [1,3,5,7,9], 'metric' : ['mahalanobis'],\n",
        "               'metric_params': [{'V': np.cov(train_x,rowvar=False)}]}]\n",
        "grid_search_cv = GridSearchCV(classifier, param_grid, cv=n_folds)\n",
        "grid_search_cv.fit(train_x,train_y)\n",
        "\n",
        "print('Combinazione migliore:\\n', grid_search_cv.best_params_)\n",
        "print('Accuratezza media della combinazione migliore: %.3f' % grid_search_cv.best_score_)\n",
        "classifier = KNeighborsClassifier(**grid_search_cv.best_params_)\n",
        "\n",
        "\n",
        "\n",
        "classifier.fit(train_x, train_y)\n",
        "\n",
        "'''mahalanobisClassifier = KNeighborsClassifier(n_neighbors=9, algorithm='brute', metric='mahalanobis',metric_params={'V': np.cov(x)})\n",
        "mahalanobisClassifier.fit(train_x,train_y)'''\n"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Combinazione migliore:\n",
            " {'metric': 'euclidean', 'n_neighbors': 9}\n",
            "Accuratezza media della combinazione migliore: 0.930\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "\"mahalanobisClassifier = KNeighborsClassifier(n_neighbors=9, algorithm='brute', metric='mahalanobis',metric_params={'V': np.cov(x)})\\nmahalanobisClassifier.fit(train_x,train_y)\""
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "id": "DtZzgq-q4CCU",
        "outputId": "29a8b6f7-0970-4a72-8e55-fbec1b140eec"
      },
      "source": [
        "pred_y_test = classifier.predict(test_x)\n",
        "pred_y_train = classifier.predict(train_x)\n",
        "'''pred_y_test_mahalanobis = mahalanobisClassifier.predict(test_x)\n",
        "pred_y_train_mahalanobis = mahalanobisClassifier.predict(train_x)'''"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'pred_y_test_mahalanobis = mahalanobisClassifier.predict(test_x)\\npred_y_train_mahalanobis = mahalanobisClassifier.predict(train_x)'"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 587
        },
        "id": "qCAbtcLRF4m3",
        "outputId": "934a4565-5897-48f3-8b3a-8dbd492fc865"
      },
      "source": [
        "print(metrics.classification_report(test_y,pred_y_test))\n",
        "print(metrics.classification_report(train_y, pred_y_train))\n",
        "\n",
        "'''print(metrics.classification_report(test_y,pred_y_test_mahalanobis))\n",
        "print(metrics.classification_report(train_y, pred_y_train_mahalanobis))'''\n",
        "\n",
        "mat = confusion_matrix(test_y, pred_y_test)\n",
        "sns.heatmap(mat.T, square=True, annot=True, fmt='d', cbar=False)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.91      0.90      0.91        80\n",
            "           1       0.95      0.95      0.95       148\n",
            "\n",
            "    accuracy                           0.93       228\n",
            "   macro avg       0.93      0.93      0.93       228\n",
            "weighted avg       0.93      0.93      0.93       228\n",
            "\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.95      0.89      0.92       132\n",
            "           1       0.94      0.97      0.95       209\n",
            "\n",
            "    accuracy                           0.94       341\n",
            "   macro avg       0.94      0.93      0.94       341\n",
            "weighted avg       0.94      0.94      0.94       341\n",
            "\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f5d5bbc1e10>"
            ]
          },
          "metadata": {},
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UnM-RgVO3I2w"
      },
      "source": [
        "scaler = StandardScaler()\n",
        "scaler.fit(train_x)\n",
        "scaled_train_x = scaler.transform(train_x)\n",
        "scaled_test_x = scaler.transform(test_x)\n"
      ],
      "execution_count": 103,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 153
        },
        "id": "xdSnrAsh44DS",
        "outputId": "447df21f-9f24-4aa3-e879-136d385c2a86"
      },
      "source": [
        "\n",
        "classifier =  KNeighborsClassifier()\n",
        "n_folds = 9\n",
        "param_grid = [{'n_neighbors' : [1,3,5,7,9], 'metric' : ['euclidean']},\n",
        "              {'n_neighbors' : [1,3,5,7,9], 'metric' : ['mahalanobis'],\n",
        "               'metric_params': [{'V': np.cov(train_x,rowvar=False)}]}]\n",
        "grid_search_cv = GridSearchCV(classifier, param_grid, cv=n_folds)\n",
        "grid_search_cv.fit(scaled_train_x,train_y)\n",
        "\n",
        "print('Combinazione migliore:\\n', grid_search_cv.best_params_)\n",
        "print('Accuratezza media della combinazione migliore: %.3f' % grid_search_cv.best_score_)\n",
        "classifier = KNeighborsClassifier(**grid_search_cv.best_params_)\n",
        "\n",
        "# Cross validation\n",
        "cv_score = cross_val_score(classifier, scaled_train_x, train_y, cv=n_folds)\n",
        "print('Accuracy di cross validation: ', cv_score)\n",
        "print('Accuracy media di cross validation: ', cv_score.mean())\n",
        "\n",
        "classifier.fit(scaled_train_x, train_y)\n",
        "\n",
        "'''mahalanobisClassifier = KNeighborsClassifier(n_neighbors=9, algorithm='brute', metric='mahalanobis',metric_params={'V': np.cov(scaled_train_x)})\n",
        "mahalanobisClassifier.fit(train_x,train_y)'''\n"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Combinazione migliore:\n",
            " {'metric': 'euclidean', 'n_neighbors': 3}\n",
            "Accuratezza media della combinazione migliore: 0.968\n",
            "Accuracy di cross validation:  [0.92105263 0.92105263 1.         0.97368421 0.94736842 1.\n",
            " 0.97368421 0.97368421 1.        ]\n",
            "Accuracy media di cross validation:  0.9678362573099414\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "\"mahalanobisClassifier = KNeighborsClassifier(n_neighbors=9, algorithm='brute', metric='mahalanobis',metric_params={'V': np.cov(scaled_train_x)})\\nmahalanobisClassifier.fit(train_x,train_y)\""
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "id": "pWak2MfK99Ce",
        "outputId": "180bdedf-145d-4f24-ea35-c3c998dc74da"
      },
      "source": [
        "pred_y_test_scaled = classifier.predict(scaled_test_x)\n",
        "pred_y_train_scaled = classifier.predict(scaled_train_x)\n",
        "'''pred_y_test_scaled_mahalanobis = mahalanobisClassifier.predict(scaled_test_x)\n",
        "pred_y_train_scaled_mahalanobis = mahalanobisClassifier.predict(scaled_train_x)'''"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'pred_y_test_scaled_mahalanobis = mahalanobisClassifier.predict(scaled_test_x)\\npred_y_train_scaled_mahalanobis = mahalanobisClassifier.predict(scaled_train_x)'"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 587
        },
        "id": "g54u_JP13EYv",
        "outputId": "c6b77efe-547c-4712-d8f0-1412a9f91e4e"
      },
      "source": [
        "print(metrics.classification_report(test_y,pred_y_test_scaled))\n",
        "print(metrics.classification_report(train_y, pred_y_train_scaled))\n",
        "'''print(metrics.classification_report(test_y,pred_y_test_scaled_mahalanobis))\n",
        "print(metrics.classification_report(train_y, pred_y_train_scaled_mahalanobis))'''\n",
        "\n",
        "mat = confusion_matrix(test_y, pred_y_test_scaled)\n",
        "sns.heatmap(mat.T, square=True, annot=True, fmt='d', cbar=False)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.95      0.91      0.93        80\n",
            "           1       0.95      0.97      0.96       148\n",
            "\n",
            "    accuracy                           0.95       228\n",
            "   macro avg       0.95      0.94      0.95       228\n",
            "weighted avg       0.95      0.95      0.95       228\n",
            "\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.98      0.96      0.97       132\n",
            "           1       0.98      0.99      0.98       209\n",
            "\n",
            "    accuracy                           0.98       341\n",
            "   macro avg       0.98      0.98      0.98       341\n",
            "weighted avg       0.98      0.98      0.98       341\n",
            "\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f5d5bb40b90>"
            ]
          },
          "metadata": {},
          "execution_count": 10
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "332LYZ8ou1eR",
        "outputId": "2b7c28d4-bd0b-4696-b8bb-90704cbbea62"
      },
      "source": [
        "sigma = np.cov(np.transpose(scaled_train_x))\n",
        "w,v = la.eig(sigma)\n",
        "w.sort()\n",
        "i = 29\n",
        "max_sum = sum(w)\n",
        "current_sum = w[29]\n",
        "ratio = current_sum/max_sum\n",
        "\n",
        "while(ratio <.95):\n",
        "  print(ratio,w[i:],i,'\\n')\n",
        "  current_sum = sum(w[i:])\n",
        "  ratio = current_sum/max_sum\n",
        "  i = i-1\n",
        "  \n",
        "\n",
        "pca = PCA(i)\n",
        "reduced_train_x = pca.fit_transform(scaled_train_x)\n",
        "print(reduced_train_x.shape,max_sum, current_sum, i)\n"
      ],
      "execution_count": 130,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0.44722202168078096 [13.45612142] 29 \n",
            "\n",
            "0.44722202168078096 [ 5.51170877 13.45612142] 28 \n",
            "\n",
            "0.6304068684862126 [ 2.98128495  5.51170877 13.45612142] 27 \n",
            "\n",
            "0.7294916077228158 [ 2.02732797  2.98128495  5.51170877 13.45612142] 26 \n",
            "\n",
            "0.7968710320264467 [ 1.61715222  2.02732797  2.98128495  5.51170877 13.45612142] 25 \n",
            "\n",
            "0.8506180267802771 [ 1.18420099  1.61715222  2.02732797  2.98128495  5.51170877 13.45612142] 24 \n",
            "\n",
            "0.8899756355233632 [ 0.69475394  1.18420099  1.61715222  2.02732797  2.98128495  5.51170877\n",
            " 13.45612142] 23 \n",
            "\n",
            "0.9130661867633284 [ 0.47080897  0.69475394  1.18420099  1.61715222  2.02732797  2.98128495\n",
            "  5.51170877 13.45612142] 22 \n",
            "\n",
            "0.9287137965525526 [ 0.42065675  0.47080897  0.69475394  1.18420099  1.61715222  2.02732797\n",
            "  2.98128495  5.51170877 13.45612142] 21 \n",
            "\n",
            "0.9426945683459902 [ 0.331726    0.42065675  0.47080897  0.69475394  1.18420099  1.61715222\n",
            "  2.02732797  2.98128495  5.51170877 13.45612142] 20 \n",
            "\n",
            "(341, 19) 30.08823529411768 28.695741980912366 19\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OmmHYSVNu8lz",
        "outputId": "29e93f4a-d039-4300-b2bd-5bdc54ba2940"
      },
      "source": [
        "classifier =  KNeighborsClassifier()\n",
        "n_folds = 9\n",
        "param_grid = [{'n_neighbors' : [1,3,5,7,9], 'metric' : ['euclidean']},\n",
        "              {'n_neighbors' : [1,3,5,7,9], 'metric' : ['mahalanobis'],\n",
        "               'metric_params': [{'V': np.cov(np.transpose(reduced_train_x))}]}]\n",
        "grid_search_cv = GridSearchCV(classifier, param_grid, cv=n_folds)\n",
        "grid_search_cv.fit(reduced_train_x,train_y)\n",
        "\n",
        "print('Combinazione migliore:\\n', grid_search_cv.best_params_)\n",
        "print('Accuratezza media della combinazione migliore: %.3f' % grid_search_cv.best_score_)\n",
        "classifier = KNeighborsClassifier(**grid_search_cv.best_params_)\n",
        "\n",
        "# Cross validation\n",
        "cv_score = cross_val_score(classifier, reduced_train_x, train_y, cv=n_folds)\n",
        "print('Accuracy di cross validation: ', cv_score)\n",
        "print('Accuracy media di cross validation: ', cv_score.mean())\n",
        "\n",
        "classifier.fit(reduced_train_x, train_y)\n"
      ],
      "execution_count": 93,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Combinazione migliore:\n",
            " {'metric': 'mahalanobis', 'metric_params': {'V': array([[510296.24548115,  -3223.77478692],\n",
            "       [ -3223.77478692,   8614.52650334]])}, 'n_neighbors': 9}\n",
            "Accuratezza media della combinazione migliore: 0.927\n",
            "Accuracy di cross validation:  [0.86842105 0.92105263 0.94736842 0.86842105 0.94736842 0.97368421\n",
            " 0.94736842 0.94736842 0.91891892]\n",
            "Accuracy media di cross validation:  0.9266635056108741\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='mahalanobis',\n",
              "                     metric_params={'V': array([[510296.24548115,  -3223.77478692],\n",
              "       [ -3223.77478692,   8614.52650334]])},\n",
              "                     n_jobs=None, n_neighbors=9, p=2, weights='uniform')"
            ]
          },
          "metadata": {},
          "execution_count": 93
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 486
        },
        "id": "PBdBKqekyicB",
        "outputId": "35c75adb-b4b1-462c-9158-b334f0040f0e"
      },
      "source": [
        "fig = plt.figure(figsize = (8,8))\n",
        "plt.scatter(reduced_x[:, 0], reduced_x[:, 1], s=50, cmap='rainbow')\n",
        "plt.show()"
      ],
      "execution_count": 101,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NE3CFZTQwNDX"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}