Bản_sao_của_Weekly_Project_MNIST.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Bản sao của Weekly-Project-MNIST.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.6"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ps5x-gYUCf9_"
      },
      "source": [
        "# Handwritten Digits Classification"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0glR10XBCxNy"
      },
      "source": [
        "## Overview\n",
        "\n",
        "The MNIST database (Modified National Institute of Standards and Technology database) is a large database of handwritten digits that is commonly used for training various image processing systems. MNIST is like the \"Hello World\" of machine learning with which you can try out a few machine learning algorithms.\n",
        "\n",
        "![](https://upload.wikimedia.org/wikipedia/commons/2/27/MnistExamples.png)\n",
        "\n",
        "Each image is 28 pixels in height and 28 pixels in width, for a total of 784 pixels in total. Each pixel has a single pixel-value associated with it, indicating the lightness or darkness of that pixel, with higher numbers meaning lighter. This pixel-value is an integer between 0 and 255, inclusive.\n",
        "\n",
        "Visually, if we omit the \"pixel\" prefix, the pixels make up the image like this:\n",
        "```\n",
        "000 001 002 003 ... 026 027\n",
        "028 029 030 031 ... 054 055\n",
        "056 057 058 059 ... 082 083\n",
        " |   |   |   |  ...  |   |\n",
        "728 729 730 731 ... 754 755\n",
        "756 757 758 759 ... 782 783 \n",
        "```"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-3hGXfqVDRna"
      },
      "source": [
        "## Get Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4IWx_-bBEYAT"
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "sns.set_style(\"whitegrid\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BNhUbXb9CSbh",
        "outputId": "26a1849d-f727-4de4-e12d-21d76b6aa7fb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        }
      },
      "source": [
        "from tensorflow.keras.datasets import mnist\n",
        "\n",
        "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
        "\n",
        "print('Training data:', X_train.shape, y_train.shape)\n",
        "print('Test data:', X_test.shape, y_test.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n",
            "Training data: (60000, 28, 28) (60000,)\n",
            "Test data: (10000, 28, 28) (10000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VYTtM9VlGs31"
      },
      "source": [
        "## EDA"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xcZURSKuDcvm"
      },
      "source": [
        "def plot_images(images, labels):\n",
        "    '''Plot images with their labels. Ten each row'''\n",
        "    plt.figure(figsize=(20, 10))\n",
        "    columns = 10\n",
        "    for i, image in enumerate(images):\n",
        "        ax = plt.subplot(len(images) / columns + 1, columns, i + 1)\n",
        "        ax.set_title(f\"{labels[i]}\", fontsize=16)\n",
        "        plt.axis('off')\n",
        "        plt.subplots_adjust(bottom=0.1)\n",
        "        plt.imshow(image, cmap='gray')\n",
        "\n",
        "def get_samples(n_samples, X, y):\n",
        "    '''Get n_samples images randomly'''\n",
        "    samples = np.random.choice(np.arange(len(y)), n_samples, replace=False)\n",
        "    return X[samples], y[samples]\n",
        "\n",
        "def get_digit(digit, n_samples, X, y):\n",
        "    '''Get n_samples images of the class digit randomly'''\n",
        "    y_digit = y[y == digit]\n",
        "    samples = np.random.choice(np.arange(len(y_digit)), n_samples, replace=False)\n",
        "    return X[y == digit][samples], y_digit[samples]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OpMe7EMCEF01",
        "outputId": "dac48c2e-9626-490a-c905-d63e5e681e70",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 436
        }
      },
      "source": [
        "n_samples = 30\n",
        "# images, labels = get_samples(n_samples, X_train, y_train)\n",
        "images, labels = get_digit(1, n_samples, X_train, y_train)\n",
        "plot_images(images, labels)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x720 with 30 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Ld_PUJxrzN1"
      },
      "source": [
        "**Plot the barplot of number of images of each class (0 to 9)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xjVYEat8EJTd",
        "outputId": "de596620-fc55-41ec-b4f3-8517fc71766d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        }
      },
      "source": [
        "# Your code here\n",
        "sns.countplot(y_test)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f0c362c8fd0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7M_TfQZUrzOG"
      },
      "source": [
        "**Print out the range of pixel values (min, max)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d0rZBlU8FmtE",
        "outputId": "b1f7be5c-32eb-4e4b-e2b8-2d6a6e871c02",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# Your code here\n",
        "print('Pixel value range:', X_train.min(), X_train.max())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Pixel value range: 0 255\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XC2sFqfcJjQ1"
      },
      "source": [
        "## Transform data for training"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6W5AWXE2rzOj"
      },
      "source": [
        "**Get the first `n_samples` images randomly from training set and test set. Then reshape them into long vectors**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XF1XGXrvJlri"
      },
      "source": [
        "n_samples = 1000\n",
        "n_test_samples = 200\n",
        "# Your code here\n",
        "X_train_flat, y_train_flat = get_samples(n_samples, X_train, y_train)\n",
        "X_test_flat, y_test_flat = get_samples(n_test_samples, X_test, y_test)\n",
        "\n",
        "# Normalization\n",
        "X_train_flat = X_train_flat.reshape((n_samples, -1))/255\n",
        "X_test_flat = X_test_flat.reshape((n_test_samples, -1))/255"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "66rNQGgnK1MX",
        "outputId": "123debfb-51d1-471b-f45d-8bf48b2e42da",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print('Training data', X_train_flat.shape)\n",
        "print('Test data', X_test_flat.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training data (1000, 784)\n",
            "Test data (200, 784)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "59B4UG39o5N1"
      },
      "source": [
        "# Get all the data\n",
        "# Run it after you finish training with the 1000 samples\n",
        "# X_train_flat = X_train.reshape((-1, 784))\n",
        "# X_test_flat = X_test.reshape((-1, 784))\n",
        "# y_train_flat = y_train\n",
        "# y_test_flat = y_test\n",
        "\n",
        "# print('Training data', X_train_flat.shape)\n",
        "# print('Test data', X_test_flat.shape)\n",
        "# print('Training label', y_train_flat.shape)\n",
        "# print('Test label', y_test_flat.shape)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bmxGMEGGLaJI"
      },
      "source": [
        "## Train models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4tJcSPEbLRsh"
      },
      "source": [
        "from sklearn.svm import SVC\n",
        "from sklearn.metrics import confusion_matrix, accuracy_score\n",
        "from sklearn.model_selection import cross_val_score\n",
        "from sklearn.model_selection import GridSearchCV"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "blM8JAmirzPb"
      },
      "source": [
        "**Train a SVM model**: Set up the training process and train with the small samples of data first before you take the full training dataset to train."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CIafedeOrzPe",
        "outputId": "8c848469-38db-453b-cc05-2366eec7c1b7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        }
      },
      "source": [
        "# Your code here\n",
        "clf = SVC(kernel='linear')\n",
        "clf.fit(X_train_flat, y_train_flat)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n",
              "    decision_function_shape='ovr', degree=3, gamma='scale', kernel='linear',\n",
              "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
              "    tol=0.001, verbose=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ENUgYnAIv7BA",
        "outputId": "b9063f9f-937f-483f-b8b4-e9d783dd3783",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        }
      },
      "source": [
        "parameters = {'gamma': [0.0001, 0.001, 0.01],\n",
        "              'C':[0.01, 0.1, 1, 10, 100]}\n",
        "\n",
        "model = SVC(kernel='rbf')\n",
        "\n",
        "gridsearch_models = GridSearchCV(estimator=model,\n",
        "                                 param_grid=parameters,\n",
        "                                 scoring='accuracy',\n",
        "                                 cv=5, n_jobs=-1)\n",
        "\n",
        "gridsearch_models.fit(X_train_flat, y_train_flat)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GridSearchCV(cv=5, error_score=nan,\n",
              "             estimator=SVC(C=1.0, break_ties=False, cache_size=200,\n",
              "                           class_weight=None, coef0=0.0,\n",
              "                           decision_function_shape='ovr', degree=3,\n",
              "                           gamma='scale', kernel='rbf', max_iter=-1,\n",
              "                           probability=False, random_state=None, shrinking=True,\n",
              "                           tol=0.001, verbose=False),\n",
              "             iid='deprecated', n_jobs=-1,\n",
              "             param_grid={'C': [0.01, 0.1, 1, 10, 100],\n",
              "                         'gamma': [0.0001, 0.001, 0.01]},\n",
              "             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,\n",
              "             scoring='accuracy', verbose=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xakJOw-uxZZZ",
        "outputId": "bb51113d-a252-43f7-bd41-c28c163f7436",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "pd.DataFrame(gridsearch_models.cv_results_)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>mean_fit_time</th>\n",
              "      <th>std_fit_time</th>\n",
              "      <th>mean_score_time</th>\n",
              "      <th>std_score_time</th>\n",
              "      <th>param_C</th>\n",
              "      <th>param_gamma</th>\n",
              "      <th>params</th>\n",
              "      <th>split0_test_score</th>\n",
              "      <th>split1_test_score</th>\n",
              "      <th>split2_test_score</th>\n",
              "      <th>split3_test_score</th>\n",
              "      <th>split4_test_score</th>\n",
              "      <th>mean_test_score</th>\n",
              "      <th>std_test_score</th>\n",
              "      <th>rank_test_score</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1.408089</td>\n",
              "      <td>0.007456</td>\n",
              "      <td>0.210397</td>\n",
              "      <td>0.005835</td>\n",
              "      <td>0.01</td>\n",
              "      <td>0.0001</td>\n",
              "      <td>{'C': 0.01, 'gamma': 0.0001}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1.438852</td>\n",
              "      <td>0.014826</td>\n",
              "      <td>0.204417</td>\n",
              "      <td>0.000865</td>\n",
              "      <td>0.01</td>\n",
              "      <td>0.001</td>\n",
              "      <td>{'C': 0.01, 'gamma': 0.001}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1.452151</td>\n",
              "      <td>0.024307</td>\n",
              "      <td>0.210251</td>\n",
              "      <td>0.005973</td>\n",
              "      <td>0.01</td>\n",
              "      <td>0.01</td>\n",
              "      <td>{'C': 0.01, 'gamma': 0.01}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1.420570</td>\n",
              "      <td>0.008494</td>\n",
              "      <td>0.215889</td>\n",
              "      <td>0.012192</td>\n",
              "      <td>0.1</td>\n",
              "      <td>0.0001</td>\n",
              "      <td>{'C': 0.1, 'gamma': 0.0001}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1.420401</td>\n",
              "      <td>0.007058</td>\n",
              "      <td>0.210699</td>\n",
              "      <td>0.005255</td>\n",
              "      <td>0.1</td>\n",
              "      <td>0.001</td>\n",
              "      <td>{'C': 0.1, 'gamma': 0.001}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>1.341415</td>\n",
              "      <td>0.008292</td>\n",
              "      <td>0.203296</td>\n",
              "      <td>0.001666</td>\n",
              "      <td>0.1</td>\n",
              "      <td>0.01</td>\n",
              "      <td>{'C': 0.1, 'gamma': 0.01}</td>\n",
              "      <td>0.615</td>\n",
              "      <td>0.580</td>\n",
              "      <td>0.595</td>\n",
              "      <td>0.590</td>\n",
              "      <td>0.610</td>\n",
              "      <td>0.598</td>\n",
              "      <td>0.012884</td>\n",
              "      <td>9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>1.426849</td>\n",
              "      <td>0.016454</td>\n",
              "      <td>0.206064</td>\n",
              "      <td>0.001690</td>\n",
              "      <td>1</td>\n",
              "      <td>0.0001</td>\n",
              "      <td>{'C': 1, 'gamma': 0.0001}</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.135</td>\n",
              "      <td>0.130</td>\n",
              "      <td>0.131</td>\n",
              "      <td>0.002000</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>1.065136</td>\n",
              "      <td>0.007661</td>\n",
              "      <td>0.198037</td>\n",
              "      <td>0.005214</td>\n",
              "      <td>1</td>\n",
              "      <td>0.001</td>\n",
              "      <td>{'C': 1, 'gamma': 0.001}</td>\n",
              "      <td>0.810</td>\n",
              "      <td>0.790</td>\n",
              "      <td>0.770</td>\n",
              "      <td>0.785</td>\n",
              "      <td>0.800</td>\n",
              "      <td>0.791</td>\n",
              "      <td>0.013565</td>\n",
              "      <td>8</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>0.725009</td>\n",
              "      <td>0.010217</td>\n",
              "      <td>0.165192</td>\n",
              "      <td>0.004452</td>\n",
              "      <td>1</td>\n",
              "      <td>0.01</td>\n",
              "      <td>{'C': 1, 'gamma': 0.01}</td>\n",
              "      <td>0.910</td>\n",
              "      <td>0.905</td>\n",
              "      <td>0.890</td>\n",
              "      <td>0.860</td>\n",
              "      <td>0.900</td>\n",
              "      <td>0.893</td>\n",
              "      <td>0.017776</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>1.048367</td>\n",
              "      <td>0.002990</td>\n",
              "      <td>0.198037</td>\n",
              "      <td>0.007238</td>\n",
              "      <td>10</td>\n",
              "      <td>0.0001</td>\n",
              "      <td>{'C': 10, 'gamma': 0.0001}</td>\n",
              "      <td>0.815</td>\n",
              "      <td>0.805</td>\n",
              "      <td>0.770</td>\n",
              "      <td>0.790</td>\n",
              "      <td>0.800</td>\n",
              "      <td>0.796</td>\n",
              "      <td>0.015297</td>\n",
              "      <td>7</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10</th>\n",
              "      <td>0.534852</td>\n",
              "      <td>0.011752</td>\n",
              "      <td>0.146861</td>\n",
              "      <td>0.001881</td>\n",
              "      <td>10</td>\n",
              "      <td>0.001</td>\n",
              "      <td>{'C': 10, 'gamma': 0.001}</td>\n",
              "      <td>0.875</td>\n",
              "      <td>0.890</td>\n",
              "      <td>0.865</td>\n",
              "      <td>0.860</td>\n",
              "      <td>0.885</td>\n",
              "      <td>0.875</td>\n",
              "      <td>0.011402</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>11</th>\n",
              "      <td>0.726112</td>\n",
              "      <td>0.018965</td>\n",
              "      <td>0.164467</td>\n",
              "      <td>0.006130</td>\n",
              "      <td>10</td>\n",
              "      <td>0.01</td>\n",
              "      <td>{'C': 10, 'gamma': 0.01}</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.925</td>\n",
              "      <td>0.915</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.905</td>\n",
              "      <td>0.012649</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>12</th>\n",
              "      <td>0.520294</td>\n",
              "      <td>0.004913</td>\n",
              "      <td>0.143874</td>\n",
              "      <td>0.001014</td>\n",
              "      <td>100</td>\n",
              "      <td>0.0001</td>\n",
              "      <td>{'C': 100, 'gamma': 0.0001}</td>\n",
              "      <td>0.880</td>\n",
              "      <td>0.880</td>\n",
              "      <td>0.870</td>\n",
              "      <td>0.855</td>\n",
              "      <td>0.875</td>\n",
              "      <td>0.872</td>\n",
              "      <td>0.009274</td>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>13</th>\n",
              "      <td>0.459755</td>\n",
              "      <td>0.006407</td>\n",
              "      <td>0.135406</td>\n",
              "      <td>0.002045</td>\n",
              "      <td>100</td>\n",
              "      <td>0.001</td>\n",
              "      <td>{'C': 100, 'gamma': 0.001}</td>\n",
              "      <td>0.875</td>\n",
              "      <td>0.905</td>\n",
              "      <td>0.885</td>\n",
              "      <td>0.885</td>\n",
              "      <td>0.875</td>\n",
              "      <td>0.885</td>\n",
              "      <td>0.010954</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14</th>\n",
              "      <td>0.725390</td>\n",
              "      <td>0.019546</td>\n",
              "      <td>0.157737</td>\n",
              "      <td>0.002808</td>\n",
              "      <td>100</td>\n",
              "      <td>0.01</td>\n",
              "      <td>{'C': 100, 'gamma': 0.01}</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.925</td>\n",
              "      <td>0.915</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.895</td>\n",
              "      <td>0.905</td>\n",
              "      <td>0.012649</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "    mean_fit_time  std_fit_time  ...  std_test_score  rank_test_score\n",
              "0        1.408089      0.007456  ...        0.002000               10\n",
              "1        1.438852      0.014826  ...        0.002000               10\n",
              "2        1.452151      0.024307  ...        0.002000               10\n",
              "3        1.420570      0.008494  ...        0.002000               10\n",
              "4        1.420401      0.007058  ...        0.002000               10\n",
              "5        1.341415      0.008292  ...        0.012884                9\n",
              "6        1.426849      0.016454  ...        0.002000               10\n",
              "7        1.065136      0.007661  ...        0.013565                8\n",
              "8        0.725009      0.010217  ...        0.017776                3\n",
              "9        1.048367      0.002990  ...        0.015297                7\n",
              "10       0.534852      0.011752  ...        0.011402                5\n",
              "11       0.726112      0.018965  ...        0.012649                1\n",
              "12       0.520294      0.004913  ...        0.009274                6\n",
              "13       0.459755      0.006407  ...        0.010954                4\n",
              "14       0.725390      0.019546  ...        0.012649                1\n",
              "\n",
              "[15 rows x 15 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DtTjXahsxpgh",
        "outputId": "c49c0808-bbfc-4550-bd70-bb9eb88ea4f8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "print('Best params', gridsearch_models.best_params_)\n",
        "print('Best score', gridsearch_models.best_score_)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Best params {'C': 10, 'gamma': 0.01}\n",
            "Best score 0.905\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yYoBam-Mx0xF"
      },
      "source": [
        "best_model = gridsearch_models.best_estimator_"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rupa4m2jX5rm"
      },
      "source": [
        "## Evaluation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UBh0Et0erzPs"
      },
      "source": [
        "**Get the predictions on test data**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1rwL5VCIX9qi",
        "outputId": "8f83adf9-3df0-4074-a9ee-678edb90298d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# Your code here\n",
        "predictions = best_model.predict(X_test_flat)\n",
        "print('Accuracy Score:', accuracy_score(y_test_flat, predictions))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy Score: 0.925\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "su0pvdthrzP9"
      },
      "source": [
        "**Plot the confusion matrix in form of a heatmap**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2RyKjfmHX7SN",
        "outputId": "2629bb45-5522-4339-fa63-1af4b4787e34",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 612
        }
      },
      "source": [
        "# Your code here\n",
        "plt.figure(figsize=(10,10))\n",
        "sns.heatmap(confusion_matrix(y_test_flat, predictions), cmap=\"YlGnBu\", annot=True, fmt=\"d\")\n",
        "plt.title(f'Accuracy score: {accuracy_score(y_test_flat, predictions)}', fontsize=20)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x720 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mpMUVV4VrzQP"
      },
      "source": [
        "**Plot the samples that the model predicts incorrectly**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-_uUXSpazDLu",
        "outputId": "85e477b8-5cbf-4bf6-f792-45c383e6219b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 450
        }
      },
      "source": [
        "# Your code here\n",
        "wrong_predicted_images = X_test_flat[y_test_flat != predictions].reshape((-1, 28, 28))\n",
        "wrong_predictions = predictions[y_test_flat != predictions]\n",
        "\n",
        "plot_images(wrong_predicted_images, wrong_predictions)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x720 with 15 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Y8O1Hh7LYNjP"
      },
      "source": [
        "## Save model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C7cNiWsLYPTE"
      },
      "source": [
        "import pickle\n",
        "\n",
        "pickle.dump(clf, open('svm_mnist.pkl', 'wb'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5B-g57XIYuH0"
      },
      "source": [
        "**Re-load model for testing**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WDsYcxfzYd8F",
        "outputId": "9796c506-d64d-4552-e095-471daf460713",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "# Your code here\n",
        "with open('svm_mnist.pkl', 'rb') as model:\n",
        "    reload_model = pickle.load(model)\n",
        "\n",
        "predictions = reload_model.predict(X_test_flat)\n",
        "print(f'Accuracy score: {accuracy_score(y_test_flat, predictions)}')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy score: 0.895\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1BVyCBz2m0Lz"
      },
      "source": [
        "## Embedding the model into a web application"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Y2hes6WnfxV"
      },
      "source": [
        "### Project structure\n",
        "\n",
        "* Clone the files of the repository\n",
        "\n",
        "```bash\n",
        "git clone https://github.com/dhminh1024/flask_doc.git\n",
        "```\n",
        "\n",
        "* Change directory\n",
        "\n",
        "```bash\n",
        "cd flask_doc\n",
        "```\n",
        "\n",
        "* Create the following folders and files\n",
        "\n",
        "|-app/\n",
        "    |-templates/\n",
        "        |-base.html\n",
        "    |-static/\n",
        "        |-js/\n",
        "            |-index.js\n",
        "        |-css/\n",
        "            |-style.css\n",
        "        |-images/\n",
        "    |-blueprints/\n",
        "        |-__ init__.py\n",
        "    |-middlewares/\n",
        "    |-models/\n",
        "    |-main.py\n",
        "\n",
        "```bash\n",
        "mkdir app app/templates app/static app/static/js app/static/css app/static/images app/blueprints app/middlewares app/models\n",
        "touch app/main.py app/templates/base.html app/static/js/index.js app/static/css/style.css\n",
        "touch app/blueprints/__init__.py\n",
        "```\n",
        "\n",
        "* Add jquery (optional)\n",
        "\n",
        "```bash\n",
        "mv ./jquery-3.4.1.min.js app/static/js/\n",
        "```\n",
        "\n",
        "* Add Canvas (optional)\n",
        "\n",
        "```bash\n",
        "mv ./index.js app/static/js/index.js\n",
        "```\n",
        "\n",
        "* Add Coderschool logo (optional)\n",
        "\n",
        "```bash\n",
        "mv ./images/coderschool-logo.png app/static/images/\n",
        "```\n",
        "\n",
        "* Add this code in **app/main.py**\n",
        "\n",
        "```python\n",
        "from flask import Flask, render_template\n",
        "from blueprints import *\n",
        "\n",
        "app = Flask(__name__)\n",
        "app.register_blueprint(home)\n",
        "\n",
        "if __name__ == '__main__':\n",
        "  app.run(host='127.0.0.1', port=5000, debug=True)\n",
        "```\n",
        "> *Hints: most of this code can be automaticaly generated with a VSCode Extension called flask-snippets by only writting \"fapp\"*\n",
        "> ![](https://i.imgur.com/npmdsha.png=500x)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oK1kJUy5rzRK"
      },
      "source": [
        "### Generating a new blueprint of our Flask app\n",
        "\n",
        "> A blueprint is a template for generating a \"section\" of our Flask app\n",
        "\n",
        "* In **flask_doc** directory, run this code in the **Terminal**\n",
        "\n",
        "```bash\n",
        "export NEW_BLUEPRINT=home\n",
        "mkdir app/blueprints/$NEW_BLUEPRINT\n",
        "touch app/blueprints/$NEW_BLUEPRINT/__init__.py app/blueprints/$NEW_BLUEPRINT/blueprint.py\n",
        "echo \"from .$NEW_BLUEPRINT import $NEW_BLUEPRINT\" >> app/blueprints/__init__.py\n",
        "echo \"from .blueprint import $NEW_BLUEPRINT\" > app/blueprints/$NEW_BLUEPRINT/__init__.py\n",
        "printf \\\n",
        "\"from flask import Blueprint, render_template, request\\n\\\n",
        "\\n\\\n",
        "$NEW_BLUEPRINT = Blueprint('$NEW_BLUEPRINT', __name__)\\\n",
        "\\n\\\n",
        "@$NEW_BLUEPRINT.route('/home')\\n\\\n",
        "def route_name():\\n\\\n",
        "    return render_template('$NEW_BLUEPRINT.html') \\n\\\n",
        "\" > app/blueprints/$NEW_BLUEPRINT/blueprint.py\n",
        "cp ./sample_page.html app/templates/$NEW_BLUEPRINT.html\n",
        "```\n",
        "> Hints: change 'home' to make a new blueprint (line 1 & 14)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "57kzglW3rzRN"
      },
      "source": [
        "### HTML Template\n",
        "\n",
        "* Add this code in **app/templates/base.html**\n",
        "\n",
        "```html\n",
        "<!doctype html>\n",
        "<html lang=\"en\">\n",
        "  <head>\n",
        "    <title>Title</title>\n",
        "    <!-- Required meta tags -->\n",
        "    <meta charset=\"utf-8\">\n",
        "    <meta name=\"viewport\" content=\"width=device-width, initial-scale=1, shrink-to-fit=no\">\n",
        "\n",
        "    <!-- Bootstrap CSS -->\n",
        "    <link rel=\"stylesheet\" href=\"https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/css/bootstrap.min.css\" integrity=\"sha384-ggOyR0iXCbMQv3Xipma34MD+dH/1fQ784/j6cY/iJTQUOhcWr7x9JvoRxT2MZw1T\" crossorigin=\"anonymous\">\n",
        "    <!-- CSS -->\n",
        "    <link href=\"static/css/style.css\" rel=\"stylesheet\">\n",
        "</head>\n",
        "  <body class=\"text-center\">\n",
        "    {% block content%} {% endblock %} \n",
        "    \n",
        "    <script src=\"static/js/jquery-3.4.1.min.js\"></script>\n",
        "    <script src=\"static/js/index.js\"></script>\n",
        "    <!-- AJAX optional -->\n",
        "    <script type=\"text/javascript\">\n",
        "        $(\"#myButton\").click(function(){\n",
        "            $('#result').text('  Predicting...');\n",
        "            var $SCRIPT_ROOT = {{request.script_root|tojson|safe}};\n",
        "            var canvasObj = document.getElementById(\"canvas\");\n",
        "            var img = canvasObj.toDataURL('image/jpeg');\n",
        "            $.ajax({\n",
        "                type: \"POST\",\n",
        "                url: $SCRIPT_ROOT + \"/upload/\",\n",
        "                data: img,\n",
        "                success: function(data){\n",
        "                    $('#result').text('Predictions ' + data);\n",
        "                }\n",
        "            });\n",
        "        });\n",
        "    </script>\n",
        "  </body>\n",
        "</html>\n",
        "```\n",
        "\n",
        "> *Hints: most of this code can be automaticaly generated with VSCode Extensions called Bootstrap 4, Font awesome 4, Font Awesome 5 Free & Pro snippets by only writting \"b4-$\"*"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UBe_IHzKrzRP"
      },
      "source": [
        "### Run the app\n",
        "\n",
        "* Change directory\n",
        "\n",
        "```bash\n",
        "cd app\n",
        "```\n",
        "\n",
        "* Run the app\n",
        "\n",
        "```bash\n",
        "python main.py\n",
        "```\n",
        "\n",
        "Congratulation, you have just created your Flask app. Here is the link of your Flask app: http://127.0.0.1:5000/home\n",
        "\n",
        "### Html for MNIST example\n",
        "\n",
        "* Replace the code of **app/templates/home.html** by this code \n",
        "\n",
        "```html\n",
        "{% extends \"base.html\" %} {% block content %}\n",
        "<div style=\"margin-top:50px\">\n",
        "  <img class=\"mb-4\" src=\"static/images/coderschool-logo.png\" alt=\"\" />\n",
        "  <h1 class=\"h3 mb-3 font-weight-normal\">Please draw a number</h1>\n",
        "\n",
        "  <canvas id=\"canvas\" width=\"400\" height=\"400\"></canvas>\n",
        "\n",
        "  <h1 class=\"h3 mb-3 font-weight-normal\" id=\"result\">Predictions:</h1>\n",
        "\n",
        "  <button id=\"myButton\" class=\"btn  btn-primary \" type=\"submit\">\n",
        "    Predict\n",
        "  </button>\n",
        "  <button id=\"clearButton\" class=\"btn btn-success\" type=\"submit\">\n",
        "    Clear\n",
        "  </button>\n",
        "\n",
        "  <p class=\"mt-5 mb-3 text-muted\">&copy; CoderSchool</p>\n",
        "</div>\n",
        "{% endblock %}\n",
        "\n",
        "```\n",
        "\n",
        "### Python for MNIST example\n",
        "\n",
        "* In **flask_doc** directory, run this code to generate the blueprint related to upload\n",
        "\n",
        "```bash\n",
        "export NEW_BLUEPRINT=upload_api\n",
        "mkdir app/blueprints/$NEW_BLUEPRINT\n",
        "touch app/blueprints/$NEW_BLUEPRINT/__init__.py app/blueprints/$NEW_BLUEPRINT/blueprint.py\n",
        "echo \"from .$NEW_BLUEPRINT import $NEW_BLUEPRINT\" >> app/blueprints/__init__.py\n",
        "echo \"from .blueprint import $NEW_BLUEPRINT\" > app/blueprints/$NEW_BLUEPRINT/__init__.py\n",
        "printf \\\n",
        "\"from flask import Blueprint, render_template, request\\n\\\n",
        "\\n\\\n",
        "$NEW_BLUEPRINT = Blueprint('$NEW_BLUEPRINT', __name__)\\\n",
        "\\n\\\n",
        "@$NEW_BLUEPRINT.route('/upload_api')\\n\\\n",
        "def route_name():\\n\\\n",
        "    return render_template('$NEW_BLUEPRINT.html') \\n\\\n",
        "\" > app/blueprints/$NEW_BLUEPRINT/blueprint.py\n",
        "cp ./sample_page.html app/templates/$NEW_BLUEPRINT.html\n",
        "```\n",
        "\n",
        "\n",
        "* Replace the code of **app/blueprints/upload_api/blueprint.py** by this code\n",
        "\n",
        "```python\n",
        "from flask import Blueprint, render_template, request\n",
        "import re\n",
        "import base64\n",
        "import numpy as np\n",
        "import pickle\n",
        "from skimage import io, color, transform\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        " \n",
        "def parse_image(imgData):\n",
        "    img_str = re.search(b\"base64,(.*)\", imgData).group(1)\n",
        "    img_decode = base64.decodebytes(img_str)\n",
        "    with open('output.png', \"wb\") as f:\n",
        "        f.write(img_decode)\n",
        "    return img_decode\n",
        "\n",
        "def preprocess_base64(base64_decoded):\n",
        "    img = io.imread(base64_decoded, plugin='imageio')\n",
        "    img = (255 - img)/255\n",
        "    img = color.rgb2gray(img)\n",
        "    img = transform.resize(img, (28,28))\n",
        "    # Testing purpose\n",
        "    # plt.imsave('processed.jpg', img, cmap='gray')\n",
        "    img = img.reshape((1, -1))\n",
        "    return img\n",
        " \n",
        "# Load your model here\n",
        "with open('models/svm_mnist.pkl', 'rb') as model:\n",
        "    model = pickle.load(model)\n",
        " \n",
        "upload_api = Blueprint('upload_api', __name__)\n",
        " \n",
        " \n",
        "@upload_api.route('/upload/', methods=['POST'])\n",
        "def upload():\n",
        "    img_decode = parse_image(request.get_data())\n",
        "    img = preprocess_base64(img_decode)\n",
        "    prediction = model.predict(img)[0]\n",
        "    return str(prediction)\n",
        "\n",
        "```\n",
        "\n",
        "* Register the upload API in **app/main.py**:\n",
        "\n",
        "```python\n",
        "from flask import Flask, render_template\n",
        "from blueprints import *\n",
        "\n",
        "app = Flask(__name__)\n",
        "app.register_blueprint(home)\n",
        "app.register_blueprint(upload_api)\n",
        "\n",
        "if __name__ == '__main__':\n",
        "    app.run(host='127.0.0.1', port=5000, debug=True)\n",
        "```"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_bBy2A8nExWB"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}