1_Gender_detection_corrigé.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "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/hocine-drioueche/OpenclassroomsProject/blob/main/1_Gender_detection_corrig%C3%A9.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Wf4Gwn3_kLuk"
      },
      "source": [
        "<center><img src=\"https://assets-datascientest.s3-eu-west-1.amazonaws.com/train/logo_datascientest.png\" style=\"height:150px;center\"></center>\n",
        "\n",
        "<hr style=\"border-width:2px;border-color:#75DFC1\">\n",
        "<center><h1>Deep-Learning - Modules complémentaires</h1></center>\n",
        "<center><h2>Classification sur des données structurées</h2></center>\n",
        "<hr style=\"border-width:2px;border-color:#75DFC1\">\n",
        "\n",
        "\n",
        "Le but de cet exercice est de s'entraîner à faire une classification sur des données structurées à l'aide d'un réseau de neurones dense.\n",
        "\n",
        " Nous allons travailler avec la base de données **```gender_classification.csv```**, qui contient les caractéristiques physiques du visage d'hommes et de femmes ainsi que leur genre.\n",
        "\n",
        "> La structure de l'exercice est la suivante :\n",
        ">> I - [Préparation du dataset](#preparation)\n",
        ">>\n",
        ">>\n",
        ">> II - [Classification avec un algorithme classique (KNN)](#knn)\n",
        ">>\n",
        ">>\n",
        ">> III - [Classification à l'aide d'un réseau de neurones dense](#dnn)\n",
        "\n",
        "- Exécuter les cellules ci-dessous pour importer le dataset et les modules nécessaires à l'exercice."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6YeEnwFD-lJN",
        "outputId": "e5a88211-01f3-41be-d62d-a7a6e7ec4a4b"
      },
      "source": [
        "!wget https://assets-datascientest.s3-eu-west-1.amazonaws.com/datasets/gender_classification.zip\n",
        "!unzip gender_classification.zip"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2021-08-03 11:59:31--  https://assets-datascientest.s3-eu-west-1.amazonaws.com/datasets/gender_classification.zip\n",
            "Resolving assets-datascientest.s3-eu-west-1.amazonaws.com (assets-datascientest.s3-eu-west-1.amazonaws.com)... 52.218.109.248\n",
            "Connecting to assets-datascientest.s3-eu-west-1.amazonaws.com (assets-datascientest.s3-eu-west-1.amazonaws.com)|52.218.109.248|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 19483 (19K) [application/zip]\n",
            "Saving to: ‘gender_classification.zip’\n",
            "\n",
            "gender_classificati 100%[===================>]  19.03K  73.7KB/s    in 0.3s    \n",
            "\n",
            "2021-08-03 11:59:33 (73.7 KB/s) - ‘gender_classification.zip’ saved [19483/19483]\n",
            "\n",
            "Archive:  gender_classification.zip\n",
            "  inflating: gender_classification.csv  \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gDQrK1n1-vQz"
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "\n",
        "from tensorflow.keras import callbacks\n",
        "from tensorflow.keras import Sequential\n",
        "from tensorflow.keras.layers import Dense, Dropout"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iJ3wpy5L-yiW"
      },
      "source": [
        "# <a name=\"preparation\"></a> I - Préparation du dataset\n",
        "\n",
        "- (a) Charger le jeu de données **```gender_classification.csv```** dans un DataFrame ```df```.\n",
        "\n",
        "- (b) Afficher la shape de ```df``` ainsi que ses 5 premières lignes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hl887xR8-yB7"
      },
      "source": [
        "df = pd.read_csv('gender_classification.csv')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 240
        },
        "id": "scrqTspoBWg6",
        "outputId": "328b0c93-f6f2-40a9-9ef4-2f45ad011470"
      },
      "source": [
        "print(\"Nombre de lignes dans le dataset : {}\".format(df.shape[0]))\n",
        "print(\"Nombre de colonnes dans le dataset : {}\".format(df.shape[1]))\n",
        "\n",
        "df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Nombre de lignes dans le dataset : 5001\n",
            "Nombre de colonnes dans le dataset : 8\n"
          ],
          "name": "stdout"
        },
        {
          "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>long_hair</th>\n",
              "      <th>forehead_width_cm</th>\n",
              "      <th>forehead_height_cm</th>\n",
              "      <th>nose_wide</th>\n",
              "      <th>nose_long</th>\n",
              "      <th>lips_thin</th>\n",
              "      <th>distance_nose_to_lip_long</th>\n",
              "      <th>gender</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>11.8</td>\n",
              "      <td>6.1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>Male</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0</td>\n",
              "      <td>14.0</td>\n",
              "      <td>5.4</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>Female</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0</td>\n",
              "      <td>11.8</td>\n",
              "      <td>6.3</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>Male</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0</td>\n",
              "      <td>14.4</td>\n",
              "      <td>6.1</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>Male</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1</td>\n",
              "      <td>13.5</td>\n",
              "      <td>5.9</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>Female</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   long_hair  forehead_width_cm  ...  distance_nose_to_lip_long  gender\n",
              "0          1               11.8  ...                          1    Male\n",
              "1          0               14.0  ...                          0  Female\n",
              "2          0               11.8  ...                          1    Male\n",
              "3          0               14.4  ...                          1    Male\n",
              "4          1               13.5  ...                          0  Female\n",
              "\n",
              "[5 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aU2_fIg1YXav"
      },
      "source": [
        "- (c) Vérifier que les colonnes de ```df``` sont du bon type, et traiter les valeurs manquantes s'il y en a."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2oCLX-uQYXlt",
        "outputId": "a33a4811-ac21-4673-80db-e3281a120422"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 5001 entries, 0 to 5000\n",
            "Data columns (total 8 columns):\n",
            " #   Column                     Non-Null Count  Dtype  \n",
            "---  ------                     --------------  -----  \n",
            " 0   long_hair                  5001 non-null   int64  \n",
            " 1   forehead_width_cm          5001 non-null   float64\n",
            " 2   forehead_height_cm         5001 non-null   float64\n",
            " 3   nose_wide                  5001 non-null   int64  \n",
            " 4   nose_long                  5001 non-null   int64  \n",
            " 5   lips_thin                  5001 non-null   int64  \n",
            " 6   distance_nose_to_lip_long  5001 non-null   int64  \n",
            " 7   gender                     5001 non-null   object \n",
            "dtypes: float64(2), int64(5), object(1)\n",
            "memory usage: 312.7+ KB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "70P6QKzlYXpt",
        "outputId": "040a9ee4-381e-409c-aedf-1b5599abbb8e"
      },
      "source": [
        "df.isna().sum()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "long_hair                    0\n",
              "forehead_width_cm            0\n",
              "forehead_height_cm           0\n",
              "nose_wide                    0\n",
              "nose_long                    0\n",
              "lips_thin                    0\n",
              "distance_nose_to_lip_long    0\n",
              "gender                       0\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Igf68u4mYBY8"
      },
      "source": [
        "- (d) Préparer les données.\n",
        "> Cette partie regroupe :\n",
        ">> - L'encodage des variables\n",
        ">>\n",
        ">>\n",
        ">> - la séparation des variables explicatives de la variable cible\n",
        ">>\n",
        ">>\n",
        ">> - La création d'un jeu de test et d'entraînement\n",
        ">>\n",
        ">>\n",
        ">> - Le scaling de ```X_train``` et ```X_test```"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xt051PgvIWHT"
      },
      "source": [
        "X = df.drop(labels='gender', axis=1)\n",
        "y = df['gender'].replace({'Male' : 0, 'Female' : 1})"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-w8hs8guIgm2"
      },
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=123)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0WwEMaa-I4nu"
      },
      "source": [
        "scaler = MinMaxScaler()\n",
        "\n",
        "scaler.fit(X_train)\n",
        "X_train_scaled = pd.DataFrame(scaler.transform(X_train), index=X_train.index, columns=X_train.columns)\n",
        "X_test_scaled = pd.DataFrame(scaler.transform(X_test), index=X_test.index, columns=X_test.columns)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mF9FgTdxI7i5"
      },
      "source": [
        "# II - <a name=\"knn\"></a> Classification avec un algorithme classique (KNN)\n",
        "\n",
        "- (a) Instancier un modèle de classification KNN.\n",
        "\n",
        "- (b) Fit le modèle aux données d'entraînement.\n",
        "\n",
        "- (c) Effectuer une prédiction à partir des données de test (attention au scaling).\n",
        "\n",
        "- (d) Afficher la matrice de confusion de la prédiction."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "T8gJGIOpI9Sm",
        "outputId": "2d4b7628-0f48-46a1-acdb-85634ba3368d"
      },
      "source": [
        "knn = KNeighborsClassifier(n_neighbors=2)\n",
        "knn.fit(X_train_scaled, y_train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
              "                     metric_params=None, n_jobs=None, n_neighbors=2, p=2,\n",
              "                     weights='uniform')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Oi2UdXp1JBR9"
      },
      "source": [
        "y_pred_knn = knn.predict(X_test_scaled)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "oz6T8JOuJExz",
        "outputId": "1013b4b9-4df9-43b2-d5d6-a8ec3dbc8231"
      },
      "source": [
        "pd.crosstab(y_test, y_pred_knn, rownames = ['Classe réelle'], colnames = ['Classe prédite par knn'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th>Classe prédite par knn</th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Classe réelle</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>505</td>\n",
              "      <td>9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>29</td>\n",
              "      <td>458</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "Classe prédite par knn    0    1\n",
              "Classe réelle                   \n",
              "0                       505    9\n",
              "1                        29  458"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "x5sQG8XEJMcc"
      },
      "source": [
        "# III - <a name=\"dnn\"></a> Classification à l'aide d'un réseau de neurones dense\n",
        "\n",
        "- (a) Instancier un réseau de neurones dense de l'architecture de votre choix.\n",
        "\n",
        "- (b) Compiler le modèle avec une métrique appropriée.\n",
        "\n",
        "- (c) Entraîner le modèle avec les paramètres de votre choix. On pourra également définir des callbacks.\n",
        "\n",
        "- (d) Effectuer une prédiction à partir des données de test (attention au scaling).\n",
        "\n",
        "- (e) Afficher la matrice de confusion de la prédiction"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kzriNHHtJPAF"
      },
      "source": [
        "input_shape = (X_train_scaled.shape[1],)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BIrOsYBSJZT2"
      },
      "source": [
        "model = Sequential()\n",
        "model.add(Dense(units=10, input_shape=input_shape, activation='relu'))\n",
        "model.add(Dense(units=128, activation='relu'))\n",
        "model.add(Dense(units=2048, activation='relu'))\n",
        "model.add(Dropout(rate=0.2))\n",
        "model.add(Dense(units=256, activation='relu'))\n",
        "model.add(Dense(units=128, activation='relu'))\n",
        "model.add(Dense(units=2, activation='softmax'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jR8prixxJerK"
      },
      "source": [
        "early_stopping = callbacks.EarlyStopping(monitor = 'val_loss',\n",
        "                        patience = 20,\n",
        "                        mode = 'min',\n",
        "                        restore_best_weights = True)\n",
        "\n",
        "lr_plateau = callbacks.ReduceLROnPlateau(monitor = 'val_loss',\n",
        "                            patience=5,\n",
        "                            factor=0.8,\n",
        "                            verbose=2,\n",
        "                            mode='min')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZnTs92E6JjAj"
      },
      "source": [
        "model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7S5XnnGfJnlz",
        "outputId": "d28b5d5c-694b-4bf6-9276-ad71e85d7e7c"
      },
      "source": [
        "history = model.fit(X_train_scaled, y_train, epochs = 200, batch_size = 30, validation_split = 0.2, callbacks = [early_stopping,lr_plateau])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/200\n",
            "107/107 [==============================] - 3s 15ms/step - loss: 0.1810 - accuracy: 0.9322 - val_loss: 0.1106 - val_accuracy: 0.9475\n",
            "Epoch 2/200\n",
            "107/107 [==============================] - 1s 14ms/step - loss: 0.1065 - accuracy: 0.9572 - val_loss: 0.1125 - val_accuracy: 0.9538\n",
            "Epoch 3/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0980 - accuracy: 0.9619 - val_loss: 0.0816 - val_accuracy: 0.9613\n",
            "Epoch 4/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0860 - accuracy: 0.9650 - val_loss: 0.0786 - val_accuracy: 0.9663\n",
            "Epoch 5/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0944 - accuracy: 0.9634 - val_loss: 0.0872 - val_accuracy: 0.9575\n",
            "Epoch 6/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0815 - accuracy: 0.9663 - val_loss: 0.0853 - val_accuracy: 0.9588\n",
            "Epoch 7/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0768 - accuracy: 0.9684 - val_loss: 0.0889 - val_accuracy: 0.9700\n",
            "Epoch 8/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0757 - accuracy: 0.9688 - val_loss: 0.0886 - val_accuracy: 0.9600\n",
            "Epoch 9/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0833 - accuracy: 0.9678 - val_loss: 0.0827 - val_accuracy: 0.9575\n",
            "\n",
            "Epoch 00009: ReduceLROnPlateau reducing learning rate to 0.000800000037997961.\n",
            "Epoch 10/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0729 - accuracy: 0.9691 - val_loss: 0.0695 - val_accuracy: 0.9700\n",
            "Epoch 11/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0703 - accuracy: 0.9716 - val_loss: 0.0857 - val_accuracy: 0.9563\n",
            "Epoch 12/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0691 - accuracy: 0.9709 - val_loss: 0.0795 - val_accuracy: 0.9600\n",
            "Epoch 13/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0760 - accuracy: 0.9688 - val_loss: 0.0768 - val_accuracy: 0.9638\n",
            "Epoch 14/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0692 - accuracy: 0.9706 - val_loss: 0.0779 - val_accuracy: 0.9638\n",
            "Epoch 15/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0682 - accuracy: 0.9712 - val_loss: 0.0741 - val_accuracy: 0.9700\n",
            "\n",
            "Epoch 00015: ReduceLROnPlateau reducing learning rate to 0.0006400000303983689.\n",
            "Epoch 16/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0635 - accuracy: 0.9728 - val_loss: 0.0760 - val_accuracy: 0.9700\n",
            "Epoch 17/200\n",
            "107/107 [==============================] - 1s 12ms/step - loss: 0.0662 - accuracy: 0.9712 - val_loss: 0.0751 - val_accuracy: 0.9675\n",
            "Epoch 18/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0664 - accuracy: 0.9731 - val_loss: 0.0814 - val_accuracy: 0.9700\n",
            "Epoch 19/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0668 - accuracy: 0.9725 - val_loss: 0.0792 - val_accuracy: 0.9663\n",
            "Epoch 20/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0668 - accuracy: 0.9734 - val_loss: 0.0800 - val_accuracy: 0.9675\n",
            "\n",
            "Epoch 00020: ReduceLROnPlateau reducing learning rate to 0.0005120000336319208.\n",
            "Epoch 21/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0621 - accuracy: 0.9734 - val_loss: 0.0779 - val_accuracy: 0.9638\n",
            "Epoch 22/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0620 - accuracy: 0.9747 - val_loss: 0.0791 - val_accuracy: 0.9688\n",
            "Epoch 23/200\n",
            "107/107 [==============================] - 1s 12ms/step - loss: 0.0588 - accuracy: 0.9750 - val_loss: 0.0926 - val_accuracy: 0.9613\n",
            "Epoch 24/200\n",
            "107/107 [==============================] - 1s 14ms/step - loss: 0.0613 - accuracy: 0.9750 - val_loss: 0.0859 - val_accuracy: 0.9663\n",
            "Epoch 25/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0600 - accuracy: 0.9747 - val_loss: 0.0842 - val_accuracy: 0.9638\n",
            "\n",
            "Epoch 00025: ReduceLROnPlateau reducing learning rate to 0.00040960004553198815.\n",
            "Epoch 26/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0588 - accuracy: 0.9753 - val_loss: 0.0817 - val_accuracy: 0.9663\n",
            "Epoch 27/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0579 - accuracy: 0.9762 - val_loss: 0.0844 - val_accuracy: 0.9650\n",
            "Epoch 28/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0577 - accuracy: 0.9769 - val_loss: 0.0855 - val_accuracy: 0.9663\n",
            "Epoch 29/200\n",
            "107/107 [==============================] - 1s 13ms/step - loss: 0.0581 - accuracy: 0.9769 - val_loss: 0.0893 - val_accuracy: 0.9663\n",
            "Epoch 30/200\n",
            "107/107 [==============================] - 1s 12ms/step - loss: 0.0558 - accuracy: 0.9762 - val_loss: 0.0863 - val_accuracy: 0.9712\n",
            "\n",
            "Epoch 00030: ReduceLROnPlateau reducing learning rate to 0.00032768002711236477.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "50OUw6hsJ9Am",
        "outputId": "8a3f021b-d560-49d5-de74-fbab63361a48"
      },
      "source": [
        "plt.figure(figsize=(12,4))\n",
        "plt.subplot(121)\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('Perte du modèle par epoch')\n",
        "plt.ylabel('perte')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='right')\n",
        "\n",
        "plt.subplot(122)\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title('Précision du modèle par epoch')\n",
        "plt.ylabel('précision')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='right');"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ux2DShXfKHVD"
      },
      "source": [
        "y_prob_dnn = model.predict(X_test_scaled)\n",
        "y_pred_dnn = y_prob_dnn.argmax(axis=-1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "id": "W2wwDxYsKRfO",
        "outputId": "44acdec9-24a5-4232-e61b-6bbbe23723d5"
      },
      "source": [
        "pd.crosstab(y_test, y_pred_dnn, rownames = ['Classe réelle'], colnames = ['Classe prédite par dnn'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th>Classe prédite par dnn</th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Classe réelle</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>502</td>\n",
              "      <td>12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>15</td>\n",
              "      <td>472</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "Classe prédite par dnn    0    1\n",
              "Classe réelle                   \n",
              "0                       502   12\n",
              "1                        15  472"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QKrYCEMj55hM"
      },
      "source": [],
      "execution_count": null,
      "outputs": []
    }
  ]
}