feature_classification_no_scale.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Zs3Vxo5dmP7j"
   },
   "source": [
    "## Feature based classification\n",
    "\n",
    "### jey@aims.ac.za"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Module importation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 101
    },
    "id": "T-5ywHOxNb7y",
    "outputId": "323ef928-249e-4493-d015-c9f7c3cde176"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<p style=\"color: red;\">\n",
       "The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n",
       "We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n",
       "or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n",
       "<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.15.0\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Conv2D, MaxPooling2D, Flatten, Dropout\n",
    "from keras.datasets import mnist\n",
    "import numpy as np\n",
    "from keras.models import model_from_json\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import mean_squared_error\n",
    "from keras.utils import np_utils\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix\n",
    "from keras.utils import to_categorical\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "%tensorflow_version 1.x\n",
    "import tensorflow as tf\n",
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "74QOVpN4WwRq"
   },
   "source": [
    "## Load the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "-f_P0_9Zhcwx"
   },
   "outputs": [],
   "source": [
    "dataset = pd.read_csv('https://drive.google.com/uc?export=download&id=1XpIaZzSMfyTUH-KFxpXFDsL1PZO_mrC0', header=None,\n",
    "                      names = ['region-centroid-col', 'region-centroid-row', 'region-pixel-count', 'short-line-density-5', \n",
    "                               'short-line-density-2', 'vedge-mean', 'vegde-sd', 'hedge-mean', 'hedge-sd', \n",
    "                               'intensity-mean', 'rawred-mean', 'rawblue-mean', 'rawgreen-mean', 'exred-mean', \n",
    "                               'exblue-mean', 'exgreen-mean', 'value-mean', 'saturatoin-mean', 'hue-mean', 'target'],delim_whitespace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0enHe6T-O4xs"
   },
   "source": [
    "## 1. Read dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 270
    },
    "id": "CqLvznMeOY0a",
    "outputId": "fdc7e14d-396b-4df1-d247-0c9eaf3f9a38"
   },
   "outputs": [
    {
     "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>region-centroid-col</th>\n",
       "      <th>region-centroid-row</th>\n",
       "      <th>region-pixel-count</th>\n",
       "      <th>short-line-density-5</th>\n",
       "      <th>short-line-density-2</th>\n",
       "      <th>vedge-mean</th>\n",
       "      <th>vegde-sd</th>\n",
       "      <th>hedge-mean</th>\n",
       "      <th>hedge-sd</th>\n",
       "      <th>intensity-mean</th>\n",
       "      <th>rawred-mean</th>\n",
       "      <th>rawblue-mean</th>\n",
       "      <th>rawgreen-mean</th>\n",
       "      <th>exred-mean</th>\n",
       "      <th>exblue-mean</th>\n",
       "      <th>exgreen-mean</th>\n",
       "      <th>value-mean</th>\n",
       "      <th>saturatoin-mean</th>\n",
       "      <th>hue-mean</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>218.0</td>\n",
       "      <td>178.0</td>\n",
       "      <td>9</td>\n",
       "      <td>0.111111</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.833333</td>\n",
       "      <td>0.547722</td>\n",
       "      <td>1.111109</td>\n",
       "      <td>0.544331</td>\n",
       "      <td>59.629630</td>\n",
       "      <td>52.444443</td>\n",
       "      <td>75.222220</td>\n",
       "      <td>51.222220</td>\n",
       "      <td>-21.555555</td>\n",
       "      <td>46.77778</td>\n",
       "      <td>-25.222221</td>\n",
       "      <td>75.222220</td>\n",
       "      <td>0.318996</td>\n",
       "      <td>-2.040554</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>113.0</td>\n",
       "      <td>130.0</td>\n",
       "      <td>9</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.277778</td>\n",
       "      <td>0.250924</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.365148</td>\n",
       "      <td>0.888889</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.555556</td>\n",
       "      <td>0.111111</td>\n",
       "      <td>-2.666667</td>\n",
       "      <td>5.00000</td>\n",
       "      <td>-2.333333</td>\n",
       "      <td>2.555556</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-2.123254</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>202.0</td>\n",
       "      <td>41.0</td>\n",
       "      <td>9</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.944448</td>\n",
       "      <td>0.772202</td>\n",
       "      <td>1.111112</td>\n",
       "      <td>1.025597</td>\n",
       "      <td>123.037040</td>\n",
       "      <td>111.888885</td>\n",
       "      <td>139.777790</td>\n",
       "      <td>117.444440</td>\n",
       "      <td>-33.444443</td>\n",
       "      <td>50.22222</td>\n",
       "      <td>-16.777779</td>\n",
       "      <td>139.777790</td>\n",
       "      <td>0.199347</td>\n",
       "      <td>-2.299918</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>32.0</td>\n",
       "      <td>173.0</td>\n",
       "      <td>9</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.722222</td>\n",
       "      <td>1.781593</td>\n",
       "      <td>9.000000</td>\n",
       "      <td>6.749488</td>\n",
       "      <td>43.592594</td>\n",
       "      <td>39.555557</td>\n",
       "      <td>52.888890</td>\n",
       "      <td>38.333336</td>\n",
       "      <td>-12.111111</td>\n",
       "      <td>27.88889</td>\n",
       "      <td>-15.777778</td>\n",
       "      <td>52.888890</td>\n",
       "      <td>0.266914</td>\n",
       "      <td>-1.998858</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>61.0</td>\n",
       "      <td>197.0</td>\n",
       "      <td>9</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.444444</td>\n",
       "      <td>1.515353</td>\n",
       "      <td>2.611111</td>\n",
       "      <td>1.925463</td>\n",
       "      <td>49.592594</td>\n",
       "      <td>44.222220</td>\n",
       "      <td>61.555557</td>\n",
       "      <td>43.000000</td>\n",
       "      <td>-16.111110</td>\n",
       "      <td>35.88889</td>\n",
       "      <td>-19.777779</td>\n",
       "      <td>61.555557</td>\n",
       "      <td>0.302925</td>\n",
       "      <td>-2.022274</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   region-centroid-col  region-centroid-row  ...  hue-mean  target\n",
       "0                218.0                178.0  ... -2.040554       6\n",
       "1                113.0                130.0  ... -2.123254       3\n",
       "2                202.0                 41.0  ... -2.299918       2\n",
       "3                 32.0                173.0  ... -1.998858       6\n",
       "4                 61.0                197.0  ... -2.022274       6\n",
       "\n",
       "[5 rows x 20 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "q7w3KipGP8PS"
   },
   "source": [
    "# 2. Split the data into training, validation and testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "y8QMPA7umk_T"
   },
   "outputs": [],
   "source": [
    "dataset[\"target\"]=dataset[\"target\"]-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "MnRN8LIyQKlW"
   },
   "outputs": [],
   "source": [
    "X = dataset.drop(columns=['target']).values\n",
    "Y = dataset['target'].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "z5ZULv-tT4Nh"
   },
   "source": [
    "### Checking the shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "Sr4S9I89QK3A",
    "outputId": "af3ef091-8a05-498c-8add-7ef782638d85"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2310, 19)"
      ]
     },
     "execution_count": 7,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "j-eIBgi_QLCO",
    "outputId": "4589a291-2574-4baa-feb3-7f8221a59694"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2310,)"
      ]
     },
     "execution_count": 8,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dXYukW4mUSyp"
   },
   "source": [
    "### Splitting the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CWn7u5DUQLFO"
   },
   "outputs": [],
   "source": [
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=.3)\n",
    "X_train, X_val, Y_train, Y_val = train_test_split(X_train, Y_train, test_size=.3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qVGW8qqAVWuR"
   },
   "source": [
    "### Checking the splitting shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "id": "cn-tCsKvVUYW",
    "outputId": "16f08592-1ac9-436d-bad7-b01896adee4e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1131, 19)\n",
      "(1131,)\n"
     ]
    }
   ],
   "source": [
    "print (X_train.shape)\n",
    "print (Y_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "id": "8t4zfENrVvYv",
    "outputId": "a31706f1-6ae7-4158-b867-6952a8a8e713"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(693, 19)\n",
      "(693,)\n"
     ]
    }
   ],
   "source": [
    "print(X_test.shape)\n",
    "print (Y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "id": "C0dSsy6hVvVL",
    "outputId": "91cdda6a-34dd-4504-fdaf-2d24957d154c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(486, 19)\n",
      "(486,)\n"
     ]
    }
   ],
   "source": [
    "print(X_val.shape)\n",
    "print (Y_val.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2t1gZbgdY6L4"
   },
   "source": [
    "### Find the unique numbers from the train labels\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 54
    },
    "id": "LJYFQXaJVvSU",
    "outputId": "57ed96e5-38fe-4583-db25-fe645051d9c2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of outputs :  7\n",
      "Output classes :  [0 1 2 3 4 5 6]\n"
     ]
    }
   ],
   "source": [
    "classes = np.unique(Y_train)\n",
    "nClasses = len(classes)\n",
    "print('Total number of outputs : ', nClasses)\n",
    "print('Output classes : ', classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ma3FYA78fAAh"
   },
   "source": [
    "### Convert from categorical labels to one-hot encoded vectors\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "hDVxH3D4VvKO"
   },
   "outputs": [],
   "source": [
    "Y_train = np_utils.to_categorical(Y_train,7)\n",
    "\n",
    "Y_val = np_utils.to_categorical(Y_val, 7)\n",
    "Y_test = np_utils.to_categorical(Y_test, 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "JxuADhJMRAB4",
    "outputId": "a5d2e848-77dc-44f6-c508-486395a4e1e2"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 1., 0., 0., 0., 0.], dtype=float32)"
      ]
     },
     "execution_count": 15,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_train[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WcfaDwRNtApQ"
   },
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "VajMENq6ZnRM",
    "outputId": "0df5d863-1b0a-43a2-c2e9-3742aa36d071"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 1., 0., 0., 0., 0.], dtype=float32)"
      ]
     },
     "execution_count": 16,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_train[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3U4u1qiLPlnN"
   },
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "BJmPfQJDZnWs",
    "outputId": "50178037-bb60-4eff-9da6-0d06220418d3"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 0., 1., 0., 0., 0.], dtype=float32)"
      ]
     },
     "execution_count": 17,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_val[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "g4ifo7OUZnai",
    "outputId": "84ec690c-6d0b-4ef2-86c2-e67040864c83"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0., 0., 1., 0., 0., 0., 0.], dtype=float32)"
      ]
     },
     "execution_count": 18,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_test[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6qVa13dLtRvl"
   },
   "source": [
    "# 4. Creation of NN model\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TWGUBI2RfoMN"
   },
   "source": [
    "### Network1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "pospSRpIZmqw"
   },
   "outputs": [],
   "source": [
    "def network1():\n",
    "  \n",
    "    model = Sequential()\n",
    "    model.add(Dense(21, input_dim=19, activation='relu'))\n",
    "    model.add(Dense(7, activation='softmax'))\n",
    "    \n",
    "    # Compile model\n",
    "    model.compile(loss='categorical_crossentropy', \n",
    "                  optimizer='adam', \n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zRBfFeGygGb8"
   },
   "source": [
    "**network1**: simple network with one layer, 19 inputs(19 features) and 21 filters between the inputs and the 7 outputs. the activation fuctions used is \"**relu**\" for the filters and \"softmax\" for the output.\n",
    "\n",
    "  Validation performance on the training was :accuracy around 65% and 75%, the confusion *matrix*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dWzaGw18fsEU"
   },
   "source": [
    "### Network2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "qaX9Gsvdfv-d"
   },
   "outputs": [],
   "source": [
    "def network2():\n",
    "  \n",
    "    model = Sequential()\n",
    "    model.add(Dense(21, input_dim=19, activation='relu'))\n",
    "    model.add(Dense(10, activation='relu'))\n",
    "\n",
    "    model.add(Dense(7, activation='softmax'))\n",
    "    \n",
    "    # Compile model\n",
    "    model.compile(loss='categorical_crossentropy', \n",
    "                  optimizer='adam', \n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U-3kcwrrgCcO"
   },
   "source": [
    "**network2**: network with 2 layers, 19 inputs(19 features), 21 filters between layer 1 and layer 2 and 10 filters between layer 2 and the 7 outputs. the activation fuctions used are \"relu\" for the both layers and \"softmax\" for the output.\n",
    "\n",
    "  Validation performance on training was :accuracy around 90% and the confusion *matrix* which has shown few prediction errors."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qUhFIexkf0_A"
   },
   "source": [
    "### Network3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "F1rk7zaDf3a2"
   },
   "outputs": [],
   "source": [
    "def network3():\n",
    "  \n",
    "    model = Sequential()\n",
    "    model.add(Dense(21, input_dim=19, activation='relu'))\n",
    "    model.add(Dense(10, activation='linear'))\n",
    "    model.add(Dense(15, activation='relu'))\n",
    "    model.add(Dense(7, activation='softmax'))\n",
    "    \n",
    "    # Compile model\n",
    "    model.compile(loss='categorical_crossentropy', \n",
    "                  optimizer='adam', \n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WO4AawHTfiK3"
   },
   "source": [
    "**network3**: network with 3 layers, 19 inputs(19 features), 21 filters between layer 1 and layer 2, 10 filters between layer 2 and layer 2, 15 filters between layer2 and layer3 and the 7 outputs. the activation fuctions used are respectively \"relu\",\"lnear\",\"relu\" for layers and \"softmax\" for the output.\n",
    "\n",
    "  Validation performance on the training was :accuracy around 80% and the confusion *matrix* which has shown more prediction errors comparative to the previous"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SZdEiQffiY0q"
   },
   "source": [
    "## Choice of model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zqbIGy2ZicX0"
   },
   "source": [
    "Based on the validation performance on the training and validation data! We decide to use the **network2** on the testing data."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vKfNBl7jgOEY"
   },
   "source": [
    "## Testing data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "jJ7aKwQpZmoB"
   },
   "outputs": [],
   "source": [
    "model = network2()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 272
    },
    "id": "I9-b7pgtZml3",
    "outputId": "b19595e0-ad9c-4ec0-d433-a3b9cb89ac92"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_10 (Dense)             (None, 21)                420       \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 10)                220       \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 7)                 77        \n",
      "=================================================================\n",
      "Total params: 717\n",
      "Trainable params: 717\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "lKKAY2_8Zmjn",
    "outputId": "1991f3d6-4db1-4371-b8c1-1dfa021a573c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1131 samples, validate on 486 samples\n",
      "Epoch 1/100\n",
      "1131/1131 [==============================] - 1s 628us/step - loss: 8.0593 - acc: 0.1813 - val_loss: 5.2039 - val_acc: 0.3333\n",
      "Epoch 2/100\n",
      "1131/1131 [==============================] - 0s 69us/step - loss: 4.9404 - acc: 0.3457 - val_loss: 3.9157 - val_acc: 0.4506\n",
      "Epoch 3/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 3.5387 - acc: 0.4598 - val_loss: 2.4733 - val_acc: 0.4753\n",
      "Epoch 4/100\n",
      "1131/1131 [==============================] - 0s 47us/step - loss: 1.9625 - acc: 0.4598 - val_loss: 1.6282 - val_acc: 0.4609\n",
      "Epoch 5/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 1.4538 - acc: 0.5544 - val_loss: 1.1535 - val_acc: 0.6049\n",
      "Epoch 6/100\n",
      "1131/1131 [==============================] - 0s 58us/step - loss: 1.1729 - acc: 0.6083 - val_loss: 1.0271 - val_acc: 0.6317\n",
      "Epoch 7/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 1.0303 - acc: 0.6428 - val_loss: 0.9265 - val_acc: 0.6605\n",
      "Epoch 8/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.9170 - acc: 0.6649 - val_loss: 0.8603 - val_acc: 0.6708\n",
      "Epoch 9/100\n",
      "1131/1131 [==============================] - 0s 62us/step - loss: 0.8248 - acc: 0.6870 - val_loss: 0.7996 - val_acc: 0.7037\n",
      "Epoch 10/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.7645 - acc: 0.7109 - val_loss: 0.7605 - val_acc: 0.7078\n",
      "Epoch 11/100\n",
      "1131/1131 [==============================] - 0s 58us/step - loss: 0.7269 - acc: 0.7233 - val_loss: 0.7483 - val_acc: 0.7058\n",
      "Epoch 12/100\n",
      "1131/1131 [==============================] - 0s 48us/step - loss: 0.6738 - acc: 0.7365 - val_loss: 0.7075 - val_acc: 0.7119\n",
      "Epoch 13/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.6705 - acc: 0.7401 - val_loss: 0.6812 - val_acc: 0.7263\n",
      "Epoch 14/100\n",
      "1131/1131 [==============================] - 0s 60us/step - loss: 0.6338 - acc: 0.7401 - val_loss: 0.6827 - val_acc: 0.7593\n",
      "Epoch 15/100\n",
      "1131/1131 [==============================] - 0s 56us/step - loss: 0.6085 - acc: 0.7515 - val_loss: 0.6582 - val_acc: 0.7243\n",
      "Epoch 16/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.5671 - acc: 0.7807 - val_loss: 0.6791 - val_acc: 0.7037\n",
      "Epoch 17/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.5674 - acc: 0.7737 - val_loss: 0.5952 - val_acc: 0.7778\n",
      "Epoch 18/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.5347 - acc: 0.7949 - val_loss: 0.5848 - val_acc: 0.7963\n",
      "Epoch 19/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.5153 - acc: 0.8126 - val_loss: 0.5609 - val_acc: 0.7798\n",
      "Epoch 20/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.5042 - acc: 0.8081 - val_loss: 0.5370 - val_acc: 0.8045\n",
      "Epoch 21/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.4988 - acc: 0.8285 - val_loss: 0.5541 - val_acc: 0.7860\n",
      "Epoch 22/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.4916 - acc: 0.8170 - val_loss: 0.5352 - val_acc: 0.8313\n",
      "Epoch 23/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.4932 - acc: 0.8267 - val_loss: 0.5634 - val_acc: 0.7901\n",
      "Epoch 24/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.4789 - acc: 0.8302 - val_loss: 0.5278 - val_acc: 0.8189\n",
      "Epoch 25/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.4293 - acc: 0.8488 - val_loss: 0.4748 - val_acc: 0.8313\n",
      "Epoch 26/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.4308 - acc: 0.8453 - val_loss: 0.4761 - val_acc: 0.8272\n",
      "Epoch 27/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.4261 - acc: 0.8585 - val_loss: 0.4663 - val_acc: 0.8313\n",
      "Epoch 28/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.4022 - acc: 0.8594 - val_loss: 0.4356 - val_acc: 0.8704\n",
      "Epoch 29/100\n",
      "1131/1131 [==============================] - 0s 56us/step - loss: 0.4183 - acc: 0.8541 - val_loss: 0.4528 - val_acc: 0.8560\n",
      "Epoch 30/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.3996 - acc: 0.8736 - val_loss: 0.4420 - val_acc: 0.8374\n",
      "Epoch 31/100\n",
      "1131/1131 [==============================] - 0s 70us/step - loss: 0.4000 - acc: 0.8444 - val_loss: 0.4848 - val_acc: 0.8313\n",
      "Epoch 32/100\n",
      "1131/1131 [==============================] - 0s 58us/step - loss: 0.3818 - acc: 0.8656 - val_loss: 0.4070 - val_acc: 0.8642\n",
      "Epoch 33/100\n",
      "1131/1131 [==============================] - 0s 56us/step - loss: 0.3613 - acc: 0.8700 - val_loss: 0.4204 - val_acc: 0.8560\n",
      "Epoch 34/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.3522 - acc: 0.8762 - val_loss: 0.4181 - val_acc: 0.8539\n",
      "Epoch 35/100\n",
      "1131/1131 [==============================] - 0s 78us/step - loss: 0.3553 - acc: 0.8638 - val_loss: 0.3971 - val_acc: 0.8601\n",
      "Epoch 36/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.3489 - acc: 0.8762 - val_loss: 0.4143 - val_acc: 0.8601\n",
      "Epoch 37/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.3372 - acc: 0.8674 - val_loss: 0.4091 - val_acc: 0.8560\n",
      "Epoch 38/100\n",
      "1131/1131 [==============================] - 0s 48us/step - loss: 0.3292 - acc: 0.8700 - val_loss: 0.3657 - val_acc: 0.8683\n",
      "Epoch 39/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.3116 - acc: 0.8789 - val_loss: 0.4272 - val_acc: 0.8457\n",
      "Epoch 40/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.3210 - acc: 0.8718 - val_loss: 0.3956 - val_acc: 0.8642\n",
      "Epoch 41/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2963 - acc: 0.8868 - val_loss: 0.3464 - val_acc: 0.8930\n",
      "Epoch 42/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.2950 - acc: 0.8842 - val_loss: 0.3372 - val_acc: 0.8909\n",
      "Epoch 43/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.2810 - acc: 0.8921 - val_loss: 0.3376 - val_acc: 0.8745\n",
      "Epoch 44/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2826 - acc: 0.8798 - val_loss: 0.3339 - val_acc: 0.8930\n",
      "Epoch 45/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2704 - acc: 0.8983 - val_loss: 0.3629 - val_acc: 0.8580\n",
      "Epoch 46/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.2696 - acc: 0.8992 - val_loss: 0.3481 - val_acc: 0.8642\n",
      "Epoch 47/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2762 - acc: 0.8930 - val_loss: 0.3204 - val_acc: 0.8704\n",
      "Epoch 48/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2632 - acc: 0.9001 - val_loss: 0.3174 - val_acc: 0.8765\n",
      "Epoch 49/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2581 - acc: 0.8983 - val_loss: 0.3549 - val_acc: 0.8601\n",
      "Epoch 50/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2637 - acc: 0.8921 - val_loss: 0.3057 - val_acc: 0.8909\n",
      "Epoch 51/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2580 - acc: 0.8966 - val_loss: 0.3132 - val_acc: 0.8704\n",
      "Epoch 52/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2534 - acc: 0.9036 - val_loss: 0.3231 - val_acc: 0.8786\n",
      "Epoch 53/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2453 - acc: 0.9027 - val_loss: 0.3212 - val_acc: 0.8724\n",
      "Epoch 54/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.2443 - acc: 0.8992 - val_loss: 0.2967 - val_acc: 0.8848\n",
      "Epoch 55/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.2446 - acc: 0.9027 - val_loss: 0.3021 - val_acc: 0.8868\n",
      "Epoch 56/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.2326 - acc: 0.9107 - val_loss: 0.2832 - val_acc: 0.8930\n",
      "Epoch 57/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.2219 - acc: 0.9125 - val_loss: 0.2803 - val_acc: 0.9136\n",
      "Epoch 58/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.2205 - acc: 0.9019 - val_loss: 0.2912 - val_acc: 0.8807\n",
      "Epoch 59/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.2167 - acc: 0.9098 - val_loss: 0.2691 - val_acc: 0.8992\n",
      "Epoch 60/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.2128 - acc: 0.9187 - val_loss: 0.2649 - val_acc: 0.8971\n",
      "Epoch 61/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.2189 - acc: 0.9072 - val_loss: 0.2701 - val_acc: 0.9033\n",
      "Epoch 62/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2185 - acc: 0.9080 - val_loss: 0.2604 - val_acc: 0.9074\n",
      "Epoch 63/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.2175 - acc: 0.9089 - val_loss: 0.2555 - val_acc: 0.9074\n",
      "Epoch 64/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.2004 - acc: 0.9151 - val_loss: 0.2568 - val_acc: 0.9012\n",
      "Epoch 65/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.2037 - acc: 0.9160 - val_loss: 0.2735 - val_acc: 0.8971\n",
      "Epoch 66/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.2016 - acc: 0.9204 - val_loss: 0.2462 - val_acc: 0.9033\n",
      "Epoch 67/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.1935 - acc: 0.9160 - val_loss: 0.2527 - val_acc: 0.9239\n",
      "Epoch 68/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.2022 - acc: 0.9204 - val_loss: 0.2617 - val_acc: 0.9115\n",
      "Epoch 69/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1974 - acc: 0.9160 - val_loss: 0.2519 - val_acc: 0.9136\n",
      "Epoch 70/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1956 - acc: 0.9142 - val_loss: 0.2374 - val_acc: 0.9095\n",
      "Epoch 71/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.1842 - acc: 0.9195 - val_loss: 0.2528 - val_acc: 0.9033\n",
      "Epoch 72/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1876 - acc: 0.9213 - val_loss: 0.2407 - val_acc: 0.9033\n",
      "Epoch 73/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1878 - acc: 0.9302 - val_loss: 0.2440 - val_acc: 0.9115\n",
      "Epoch 74/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.1847 - acc: 0.9213 - val_loss: 0.2490 - val_acc: 0.9053\n",
      "Epoch 75/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1959 - acc: 0.9151 - val_loss: 0.2343 - val_acc: 0.9136\n",
      "Epoch 76/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1706 - acc: 0.9284 - val_loss: 0.2292 - val_acc: 0.9198\n",
      "Epoch 77/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.1784 - acc: 0.9187 - val_loss: 0.2293 - val_acc: 0.9095\n",
      "Epoch 78/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.1822 - acc: 0.9187 - val_loss: 0.2355 - val_acc: 0.9074\n",
      "Epoch 79/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1669 - acc: 0.9302 - val_loss: 0.2604 - val_acc: 0.9033\n",
      "Epoch 80/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.1701 - acc: 0.9302 - val_loss: 0.2480 - val_acc: 0.9095\n",
      "Epoch 81/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1746 - acc: 0.9310 - val_loss: 0.2365 - val_acc: 0.9095\n",
      "Epoch 82/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1696 - acc: 0.9222 - val_loss: 0.2297 - val_acc: 0.9095\n",
      "Epoch 83/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1784 - acc: 0.9195 - val_loss: 0.2237 - val_acc: 0.9218\n",
      "Epoch 84/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1675 - acc: 0.9319 - val_loss: 0.2132 - val_acc: 0.9177\n",
      "Epoch 85/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.1635 - acc: 0.9355 - val_loss: 0.2152 - val_acc: 0.9218\n",
      "Epoch 86/100\n",
      "1131/1131 [==============================] - 0s 49us/step - loss: 0.1593 - acc: 0.9310 - val_loss: 0.2220 - val_acc: 0.9033\n",
      "Epoch 87/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1605 - acc: 0.9310 - val_loss: 0.2318 - val_acc: 0.9198\n",
      "Epoch 88/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1589 - acc: 0.9443 - val_loss: 0.2036 - val_acc: 0.9239\n",
      "Epoch 89/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.1518 - acc: 0.9355 - val_loss: 0.1977 - val_acc: 0.9280\n",
      "Epoch 90/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1478 - acc: 0.9399 - val_loss: 0.1965 - val_acc: 0.9259\n",
      "Epoch 91/100\n",
      "1131/1131 [==============================] - 0s 55us/step - loss: 0.1617 - acc: 0.9319 - val_loss: 0.1941 - val_acc: 0.9321\n",
      "Epoch 92/100\n",
      "1131/1131 [==============================] - 0s 54us/step - loss: 0.1477 - acc: 0.9425 - val_loss: 0.2133 - val_acc: 0.9115\n",
      "Epoch 93/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1480 - acc: 0.9408 - val_loss: 0.2037 - val_acc: 0.9218\n",
      "Epoch 94/100\n",
      "1131/1131 [==============================] - 0s 50us/step - loss: 0.1417 - acc: 0.9452 - val_loss: 0.1770 - val_acc: 0.9342\n",
      "Epoch 95/100\n",
      "1131/1131 [==============================] - 0s 52us/step - loss: 0.1406 - acc: 0.9461 - val_loss: 0.1819 - val_acc: 0.9259\n",
      "Epoch 96/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1410 - acc: 0.9443 - val_loss: 0.1766 - val_acc: 0.9300\n",
      "Epoch 97/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1452 - acc: 0.9514 - val_loss: 0.2192 - val_acc: 0.9218\n",
      "Epoch 98/100\n",
      "1131/1131 [==============================] - 0s 56us/step - loss: 0.1688 - acc: 0.9248 - val_loss: 0.1945 - val_acc: 0.9218\n",
      "Epoch 99/100\n",
      "1131/1131 [==============================] - 0s 53us/step - loss: 0.1386 - acc: 0.9434 - val_loss: 0.1795 - val_acc: 0.9259\n",
      "Epoch 100/100\n",
      "1131/1131 [==============================] - 0s 51us/step - loss: 0.1302 - acc: 0.9549 - val_loss: 0.1661 - val_acc: 0.9362\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, Y_train, validation_data=(X_val, Y_val) ,epochs = 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sM9XFoqtvMZf"
   },
   "outputs": [],
   "source": [
    "prediction = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 254
    },
    "id": "yvLhUQtsvMjb",
    "outputId": "d440afd3-3ca9-4708-f47e-4bcf812c15a5"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.0061443e-03, 2.0192775e-11, 4.8491216e-01, ..., 1.7399164e-01,\n",
       "        5.2535376e-10, 9.3826662e-09],\n",
       "       [9.5882267e-01, 1.0237470e-09, 3.7208906e-06, ..., 3.1379573e-02,\n",
       "        1.8549887e-03, 2.5342119e-05],\n",
       "       [6.6144744e-06, 3.3840422e-09, 1.0450917e-06, ..., 1.3307115e-03,\n",
       "        9.8743314e-01, 1.5231285e-03],\n",
       "       ...,\n",
       "       [9.9985933e-01, 2.2983121e-11, 5.7288157e-11, ..., 1.5520289e-06,\n",
       "        3.4878917e-06, 2.4739056e-12],\n",
       "       [2.9282823e-05, 8.9202874e-04, 1.9462923e-10, ..., 1.2551819e-05,\n",
       "        8.0066047e-07, 3.0035185e-15],\n",
       "       [1.2237915e-20, 9.9430048e-01, 1.4552989e-17, ..., 9.1762216e-05,\n",
       "        2.4053730e-19, 1.3028259e-24]], dtype=float32)"
      ]
     },
     "execution_count": 98,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "rrHg8gDwvMnu"
   },
   "outputs": [],
   "source": [
    "prediction_classes = model.predict_classes(X_test)\n",
    "prediction_classes_val = model.predict_classes(X_val)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 599
    },
    "id": "GHkUGMWovMgM",
    "outputId": "6928d11c-454b-44d3-f397-be846ad54e1b"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2, 0, 5, 3, 3, 2, 4, 6, 5, 3, 1, 5, 1, 2, 5, 3, 4, 5, 5, 6, 5, 1,\n",
       "       1, 5, 3, 0, 3, 5, 4, 6, 5, 3, 0, 2, 2, 6, 4, 6, 5, 0, 0, 5, 5, 3,\n",
       "       3, 0, 6, 0, 2, 3, 5, 1, 5, 3, 2, 0, 0, 1, 1, 0, 0, 1, 0, 4, 2, 1,\n",
       "       4, 1, 4, 1, 1, 3, 2, 2, 1, 4, 5, 5, 4, 3, 3, 6, 1, 2, 1, 3, 2, 3,\n",
       "       1, 0, 0, 2, 6, 2, 3, 0, 3, 5, 1, 3, 1, 5, 1, 5, 4, 2, 2, 0, 3, 3,\n",
       "       4, 6, 5, 6, 1, 1, 5, 5, 4, 6, 2, 5, 2, 1, 5, 5, 5, 0, 2, 2, 4, 3,\n",
       "       2, 5, 2, 4, 5, 5, 0, 0, 4, 1, 5, 1, 0, 1, 4, 3, 2, 6, 0, 0, 6, 3,\n",
       "       1, 1, 3, 3, 2, 2, 1, 5, 3, 0, 4, 5, 5, 0, 5, 0, 6, 1, 4, 4, 0, 6,\n",
       "       0, 0, 5, 1, 4, 3, 3, 5, 5, 4, 5, 4, 3, 1, 5, 3, 0, 5, 6, 0, 4, 5,\n",
       "       2, 6, 2, 4, 3, 3, 2, 2, 5, 2, 0, 6, 3, 1, 1, 2, 5, 2, 5, 1, 6, 4,\n",
       "       5, 6, 0, 3, 5, 3, 2, 1, 4, 2, 6, 4, 1, 1, 4, 1, 2, 1, 2, 0, 3, 3,\n",
       "       0, 6, 5, 2, 0, 5, 0, 2, 4, 3, 2, 3, 6, 3, 1, 1, 3, 0, 3, 2, 5, 2,\n",
       "       1, 3, 6, 5, 0, 2, 5, 2, 3, 1, 1, 2, 2, 3, 3, 3, 3, 6, 1, 5, 0, 6,\n",
       "       5, 6, 1, 1, 1, 3, 3, 0, 3, 6, 3, 3, 1, 2, 1, 5, 5, 2, 0, 4, 6, 2,\n",
       "       0, 3, 3, 2, 0, 2, 5, 0, 2, 4, 5, 4, 1, 4, 2, 3, 0, 6, 1, 2, 6, 3,\n",
       "       3, 4, 5, 3, 0, 6, 5, 2, 2, 2, 0, 6, 2, 6, 6, 0, 5, 4, 5, 4, 4, 2,\n",
       "       0, 6, 3, 2, 0, 0, 4, 5, 0, 6, 6, 4, 3, 2, 5, 6, 2, 5, 3, 4, 6, 4,\n",
       "       2, 2, 6, 2, 3, 1, 5, 4, 5, 6, 3, 1, 6, 0, 6, 3, 3, 2, 3, 1, 6, 3,\n",
       "       1, 2, 3, 6, 2, 3, 3, 6, 2, 6, 4, 5, 5, 4, 0, 1, 1, 4, 0, 4, 1, 5,\n",
       "       0, 3, 3, 3, 3, 3, 4, 3, 0, 3, 2, 6, 4, 6, 1, 4, 6, 6, 4, 6, 2, 6,\n",
       "       3, 4, 6, 1, 0, 1, 2, 5, 6, 1, 0, 4, 3, 4, 5, 0, 1, 6, 6, 6, 0, 5,\n",
       "       6, 1, 1, 6, 2, 1, 5, 1, 3, 0, 5, 4, 2, 0, 6, 3, 4, 1, 2, 6, 6, 6,\n",
       "       5, 5, 1, 5, 1, 0, 6, 0, 6, 5, 2, 3, 0, 5, 0, 0, 1, 0, 0, 5, 0, 0,\n",
       "       1, 3, 0, 4, 1, 2, 2, 6, 1, 1, 3, 5, 3, 0, 0, 6, 4, 2, 2, 3, 4, 5,\n",
       "       4, 2, 0, 6, 2, 0, 1, 1, 6, 1, 0, 1, 2, 2, 1, 1, 0, 4, 4, 2, 2, 1,\n",
       "       0, 4, 6, 4, 5, 4, 0, 3, 4, 2, 3, 0, 1, 2, 3, 3, 6, 6, 3, 6, 3, 6,\n",
       "       1, 0, 6, 5, 2, 1, 5, 3, 3, 5, 2, 5, 6, 3, 0, 3, 2, 1, 6, 4, 5, 5,\n",
       "       4, 3, 6, 1, 0, 4, 3, 6, 0, 3, 0, 3, 6, 2, 3, 0, 2, 5, 2, 5, 2, 0,\n",
       "       2, 0, 5, 0, 1, 6, 0, 6, 3, 3, 6, 5, 0, 1, 5, 2, 2, 5, 5, 2, 2, 4,\n",
       "       0, 0, 4, 0, 0, 6, 5, 4, 6, 5, 2, 4, 0, 1, 0, 6, 1, 2, 2, 4, 4, 6,\n",
       "       4, 3, 3, 0, 6, 0, 1, 4, 2, 0, 6, 3, 0, 4, 5, 1, 3, 3, 5, 4, 2, 3,\n",
       "       2, 6, 6, 0, 2, 2, 4, 1, 0, 3, 1])"
      ]
     },
     "execution_count": 100,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction_classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 145
    },
    "id": "y7YXBhF0vMeK",
    "outputId": "57239e37-8074-46ff-d65d-fce75bcc638e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 1., ..., 0., 0., 0.],\n",
       "       [1., 0., 0., ..., 0., 0., 0.],\n",
       "       [0., 0., 0., ..., 0., 1., 0.],\n",
       "       ...,\n",
       "       [1., 0., 0., ..., 0., 0., 0.],\n",
       "       [0., 0., 0., ..., 0., 0., 0.],\n",
       "       [0., 1., 0., ..., 0., 0., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 101,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_test"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U8ttCIdUjNvw"
   },
   "source": [
    "# 5) Plot the confusion matrix and describe two of the errors which were made by the model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6E9EF3k9jsp9"
   },
   "source": [
    "### Confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 145
    },
    "id": "VOTkHysOvMXW",
    "outputId": "048e3e4c-014a-4a39-927d-d04f912ab718"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[105,   0,   0,   1,   3,   0,   0],\n",
       "       [  0,  96,   0,   0,   0,   0,   0],\n",
       "       [  0,   0,  97,   4,   2,   0,   0],\n",
       "       [  0,   0,   0,  93,   7,   6,   0],\n",
       "       [  0,   0,  10,  13,  67,   0,   0],\n",
       "       [  0,   0,   0,   1,   0,  95,   0],\n",
       "       [  0,   0,   0,   0,   0,   0,  93]])"
      ]
     },
     "execution_count": 102,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(np.argmax(Y_test,1), prediction_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Tt3C_XTljwAy"
   },
   "source": [
    "## Mean square errors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Hjo3tu-FkbzL"
   },
   "outputs": [],
   "source": [
    "val_prediction = model.predict(X_val)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JasUUsKWk3gn"
   },
   "source": [
    "### For the validation\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "gDuwrNPjj3c0",
    "outputId": "f3029d82-c332-4be2-9fae-ad81752ea173"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.013435135"
      ]
     },
     "execution_count": 104,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_squared_error(Y_val, val_prediction)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FgCf7gANk9Zo"
   },
   "source": [
    "### For the prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "XklyxN_XksAD",
    "outputId": "9a59f78d-5b1a-470e-8419-fd613c8239df"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0142778745"
      ]
     },
     "execution_count": 105,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_squared_error(Y_test,prediction)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UrsHDR1ly6ll"
   },
   "source": [
    "# 6. Accuracy\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "APpWEDj2lVbp"
   },
   "source": [
    "### For the test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "SbTwLDynVu7l",
    "outputId": "12811dbc-fa27-45df-f4f8-32a6c03256a7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9321789321789322"
      ]
     },
     "execution_count": 106,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(np.argmax(Y_test,1), prediction_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "r96IVOMhWR8o"
   },
   "source": [
    "### For the validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "im1EwfcaWWLx",
    "outputId": "3a60514f-6154-48e6-ef9e-970cbe48e088"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9362139917695473"
      ]
     },
     "execution_count": 107,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(np.argmax(Y_val,1), prediction_classes_val)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DFArDoYszFz4"
   },
   "source": [
    "# 7. Plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "e_z3YDsMy_cK"
   },
   "outputs": [],
   "source": [
    "def plot_hist(h, xsize=6, ysize=10):\n",
    "\n",
    "    fig_size = plt.rcParams[\"figure.figsize\"]\n",
    "    plt.rcParams[\"figure.figsize\"] = [xsize, ysize]\n",
    "    fig, axes = plt.subplots(nrows=1, ncols=1, sharex=True)\n",
    "    \n",
    "    # summarize history for Accuracy\n",
    "    plt.subplot(211)\n",
    "    plt.plot(h['acc'])\n",
    "    plt.plot(h['val_acc'])\n",
    "    plt.title('Training Performance')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.legend(['Train', 'Validation'], loc='best')\n",
    "    \n",
    "    plt.draw()\n",
    "    plt.show()\n",
    "\n",
    "    return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 374
    },
    "id": "WfLl08Y3y_qA",
    "outputId": "63e8bf3c-773c-4eee-a78a-1c40283b6656"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfEAAAFlCAYAAAD7xdEoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDfX/AEI01a7b5Eu8AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 576x864 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_hist(history.history, xsize=8, ysize=12)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RS3eiMtwY0vN"
   },
   "source": [
    "### Conclusion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DzlEphtGY7Ux"
   },
   "source": [
    "We have an overfitting! Because the predicticting data is perfectly fit in the training . Another reason could be the meansquare errors small"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KkFzh6Ux2tHO"
   },
   "source": [
    "# 9. Saving the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HOFzRt6r29Os"
   },
   "source": [
    "### Save the model to the local disk\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "iwPWfEBBy_tB"
   },
   "outputs": [],
   "source": [
    "model_json = model.to_json()\n",
    "with open(\"model.json\", \"w\") as json_file:\n",
    "    json_file.write(model_json)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "GvLKwgYF3BEO"
   },
   "source": [
    "## Save the model's weights to the local disk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Lc4FR4IMy_y_"
   },
   "outputs": [],
   "source": [
    "model.save_weights(\"model_weights.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Rl4IeeWh3JjK"
   },
   "source": [
    "### Load the model back into memory\n",
    "\n",
    "We first read in the JSON file then make use of *model_from_json* which we obtained as an import at the top of this script: *from keras.models import model_from_json*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ZM2-YkP4y_3F"
   },
   "outputs": [],
   "source": [
    "json_file = open('model.json', 'r')\n",
    "loaded_model_as_json = json_file.read()\n",
    "json_file.close()\n",
    "restored_model = model_from_json(loaded_model_as_json)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "iPA91SNc3QrC"
   },
   "source": [
    "## Load the weights back into memory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "_awJcWWyy_wi"
   },
   "outputs": [],
   "source": [
    "restored_model.load_weights(\"model_weights.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U_YBmA0b3XFM"
   },
   "source": [
    "## Re-compile the model!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "GQhqalEf3cq7"
   },
   "outputs": [],
   "source": [
    "restored_model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9RG0ABdG3fuS"
   },
   "source": [
    "## Resume training, predict or perform other tasks\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 126
    },
    "id": "TTBKCPmO3l_4",
    "outputId": "f2f6745e-941b-4c3e-80df-c6b9049fdd31"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1131 samples, validate on 486 samples\n",
      "Epoch 1/2\n",
      "1131/1131 [==============================] - 1s 776us/step - loss: 0.0529 - acc: 0.9769 - val_loss: 0.0535 - val_acc: 0.9750\n",
      "Epoch 2/2\n",
      "1131/1131 [==============================] - 0s 195us/step - loss: 0.0529 - acc: 0.9779 - val_loss: 0.0477 - val_acc: 0.9788\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7ff1a095add8>"
      ]
     },
     "execution_count": 115,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "restored_model.fit(X_train, Y_train, validation_data=(X_val, Y_val), epochs=2, batch_size=8, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CHc2akq93tn0"
   },
   "outputs": [],
   "source": [
    "prediction_classes_val_restored = restored_model.predict_classes(X_val)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U-82d-1Z6jwb"
   },
   "source": [
    "## Check the accuracy of the restored model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "-xl36i2E3yxP",
    "outputId": "9befbe3b-4a03-47b8-b9ca-59bf279a9eae"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9259259259259259"
      ]
     },
     "execution_count": 117,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(np.argmax(Y_val,1), prediction_classes_val_restored)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-_ZTt9w03pzp"
   },
   "source": [
    "## Notice\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "s7IVvJvv360v"
   },
   "source": [
    "We notice that the accuracy of the restored model is approximatively the same as the accuracy of the model before saving.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "k36fxpt9tVbh"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "jérémie_bakambana_assign_1_no_scale.ipynb",
   "provenance": []
  },
  "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}