uci_har_lstm.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "uci-har-lstm.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "A5GD1ldUeXgt",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#!pip install tensorflow==1.4\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import os\n",
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import scipy.io as sio\n",
        "from IPython.display import display\n",
        "import matplotlib.pyplot as plt\n",
        "import pywt\n",
        "import scipy.stats\n",
        "\n",
        "import datetime as dt\n",
        "from collections import defaultdict, Counter\n",
        "#import tensorflow as tf  # Version 1.0.0 (some previous versions are used in past commits)\n",
        "from sklearn import metrics\n",
        "\n",
        "\n",
        "import os"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kltqwByl4Wr2",
        "colab_type": "text"
      },
      "source": [
        "## Reading UCI-HAR"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T0Lw2cJbf9zP",
        "colab_type": "code",
        "outputId": "49870dcc-0d42-4c7d-e037-427e8de43fa7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 121
        }
      },
      "source": [
        "activities_description = {\n",
        "    1: 'walking',\n",
        "    2: 'walking upstairs',\n",
        "    3: 'walking downstairs',\n",
        "    4: 'sitting',\n",
        "    5: 'standing',\n",
        "    6: 'laying'\n",
        "}\n",
        "\n",
        "def signal_reader(filename):\n",
        "    with open(filename, 'r') as fp:\n",
        "        d = fp.read().splitlines()\n",
        "        d = map(lambda x: x.rstrip().lstrip().split(), d)\n",
        "        d = [list(map(float, line)) for line in d]\n",
        "    return d\n",
        "\n",
        "def label_reader(filename):        \n",
        "    with open(filename, 'r') as fp:\n",
        "        act = fp.read().splitlines()\n",
        "        act = list(map(int, act))\n",
        "    return act\n",
        "\n",
        "def randomizer(dataset, labels):\n",
        "    p = np.random.permutation(labels.shape[0])\n",
        "    shuffled_dataset = dataset[p, :, :]\n",
        "    shuffled_labels = labels[p]\n",
        "    return shuffled_dataset, shuffled_labels\n",
        "\n",
        "train_fold = '/content/drive/My Drive/UCI HAR Dataset-Full/train/Inertial Signals/'\n",
        "test_fold = '/content/drive/My Drive/UCI HAR Dataset-Full/test/Inertial Signals/'\n",
        "\n",
        "train = ['body_acc_x_train.txt', 'body_acc_y_train.txt', 'body_acc_z_train.txt', \n",
        "                     'body_gyro_x_train.txt', 'body_gyro_y_train.txt', 'body_gyro_z_train.txt',\n",
        "                     'total_acc_x_train.txt', 'total_acc_y_train.txt', 'total_acc_z_train.txt']\n",
        "\n",
        "test = ['body_acc_x_test.txt', 'body_acc_y_test.txt', 'body_acc_z_test.txt', \n",
        "                     'body_gyro_x_test.txt', 'body_gyro_y_test.txt', 'body_gyro_z_test.txt',\n",
        "                     'total_acc_x_test.txt', 'total_acc_y_test.txt', 'total_acc_z_test.txt']\n",
        "\n",
        "train_y_files = '/content/drive/My Drive/UCI HAR Dataset-Full/train/y_train.txt'\n",
        "test_y_files = '/content/drive/My Drive/UCI HAR Dataset-Full/test/y_test.txt'\n",
        "\n",
        "arr_train_signals, arr_test_signals = [], []\n",
        "\n",
        "for inp_file in train:\n",
        "    signal = signal_reader(train_fold + inp_file)\n",
        "    arr_train_signals.append(signal)\n",
        "arr_train_signals = np.transpose(np.array(arr_train_signals), (1, 2, 0))\n",
        "\n",
        "for inp_file in test:\n",
        "    signal = signal_reader(test_fold + inp_file)\n",
        "    arr_test_signals.append(signal)\n",
        "arr_test_signals = np.transpose(np.array(arr_test_signals), (1, 2, 0))\n",
        "\n",
        "labels_train = label_reader(train_y_files)\n",
        "labels_test = label_reader(test_y_files)\n",
        "\n",
        "[no_signals_train, no_steps_train, no_components_train] = np.shape(arr_train_signals)\n",
        "[no_signals_test, no_steps_test, no_components_test] = np.shape(arr_train_signals)\n",
        "no_labels = len(np.unique(labels_train[:]))\n",
        "\n",
        "print(\"The train  has {} signals, each one of length {} and {} components \".format(no_signals_train, no_steps_train, no_components_train))\n",
        "print(\"The test  has {} signals, each one of length {} and {} components \".format(no_signals_test, no_steps_test, no_components_test))\n",
        "print(\"The train  has {} labels, with the following distribution:\\n {}\".format(np.shape(labels_train)[0], Counter(labels_train[:])))\n",
        "print(\"The test  has {} labels, with the following distribution:\\n {}\".format(np.shape(labels_test)[0], Counter(labels_test[:])))\n",
        "\n",
        "x_train, y_train = randomizer(arr_train_signals, np.array(labels_train))\n",
        "x_test, y_test = randomizer(arr_test_signals, np.array(labels_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "The train  has 7352 signals, each one of length 128 and 9 components \n",
            "The test  has 7352 signals, each one of length 128 and 9 components \n",
            "The train  has 7352 labels, with the following distribution:\n",
            " Counter({6: 1407, 5: 1374, 4: 1286, 1: 1226, 2: 1073, 3: 986})\n",
            "The test  has 2947 labels, with the following distribution:\n",
            " Counter({6: 537, 5: 532, 1: 496, 4: 491, 2: 471, 3: 420})\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uV-c-MhW2b61",
        "colab_type": "code",
        "outputId": "696f4215-3200-4ce2-a2bc-d00286dd53ee",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        " len(arr_train_signals[0])  # 128 timesteps per series\n",
        " len(arr_train_signals[0][0])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "9"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pMfyGkYawmLJ",
        "colab_type": "code",
        "outputId": "5362adb4-2efc-44f0-e2c7-63e85f99cb50",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "print(np.array(arr_train_signals).shape)\n",
        "print(np.array(arr_test_signals).shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(7352, 128, 9)\n",
            "(2947, 128, 9)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PYly09ba6K2F",
        "colab_type": "text"
      },
      "source": [
        "## LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-cbtAk7k62GP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import keras\n",
        "from keras.layers import Dense, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras.models import Sequential\n",
        "from keras.callbacks import History,EarlyStopping \n",
        "from sklearn.metrics import classification_report\n",
        "from sklearn.metrics import f1_score\n",
        "from keras import backend as K\n",
        "\n",
        "\n",
        "\n",
        "def recall_m(true_Y, pred_Y):\n",
        "        TP = K.sum(K.round(K.clip(true_Y * pred_Y, 0, 1)))\n",
        "        possible_pos = K.sum(K.round(K.clip(true_Y, 0, 1)))\n",
        "        rec = TP / (possible_pos + K.epsilon())\n",
        "        return rec\n",
        "\n",
        "def precision_m(true_Y, pred_Y):\n",
        "        true_positives = K.sum(K.round(K.clip(true_Y * pred_Y, 0, 1)))\n",
        "        predicted_positives = K.sum(K.round(K.clip(pred_Y, 0, 1)))\n",
        "        precision = true_positives / (predicted_positives + K.epsilon())\n",
        "        return precision\n",
        "\n",
        "def f1_m(true_Y, pred_Y):\n",
        "    pres = precision_m(true_Y, pred_Y)\n",
        "    rec = recall_m(true_Y, pred_Y)\n",
        "    return 2*((pres*rec)/(pres+rec+K.epsilon()))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SZvzEWti6HOy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# calısan\n",
        "\n",
        "\n",
        "from keras.layers import LSTM,Lambda\n",
        "from keras import Model\n",
        "\n",
        "\n",
        "class lstm(object):\n",
        "  def __init__(self,\n",
        "               class_num=6,\n",
        "               last_activation='softmax'):\n",
        "\n",
        "      self.class_num = 6\n",
        "      self.last_activation = last_activation\n",
        "\n",
        "  def forward(self):\n",
        "    \n",
        "      img_inputs  = keras.Input((128,9))\n",
        "      \n",
        "      print(img_inputs.shape)\n",
        "      \n",
        "      \n",
        "\n",
        "      \n",
        "      x=LSTM(64, return_sequences=False)(img_inputs) \n",
        "      print(x.shape)\n",
        "\n",
        "\n",
        "      x = keras.layers.Dense(128, activation='relu')(x)\n",
        "      print(x.shape)\n",
        "\n",
        "\n",
        "      output = Dense(self.class_num, activation=self.last_activation)(x)\n",
        "      print(output.shape)\n",
        "\n",
        "      model = Model(inputs=img_inputs, outputs=output)\n",
        "\n",
        "\n",
        "      return model\n",
        "\n",
        "\n",
        "def lstm_train(X_train,y_train,X_test,y_test):\n",
        "  \n",
        "  history = History()\n",
        "\n",
        "  num_classes = 6\n",
        "  batch_size = 64\n",
        "  epochs = 50\n",
        "  early_stopping = [EarlyStopping(monitor='val_loss',min_delta=0,restore_best_weights=True, patience=7,verbose=1, mode='auto')]\n",
        " \n",
        "\n",
        "\n",
        "  model=lstm().forward()\n",
        "  model.compile(loss=keras.losses.categorical_crossentropy, \n",
        "                optimizer=keras.optimizers.Adam(lr=0.01), \n",
        "                metrics=['accuracy',f1_m,precision_m, recall_m])\n",
        "\n",
        "  history=model.fit(X_train, y_train, batch_size=batch_size, \n",
        "            epochs=epochs, verbose=1, \n",
        "            validation_split=0.1, \n",
        "            callbacks=early_stopping)\n",
        "\n",
        "  \n",
        "\n",
        "  loss, accuracy, f1_score, precision, recall = model.evaluate(X_train, y_train, verbose=1)\n",
        "  print(\" CNN Training Loss: {:.4f}\".format(loss))\n",
        "  print(\" CNN Training Accuracy: {:.4f}\".format(accuracy))\n",
        "  print(\" CNN Training f1 score: {:.4f}\".format(f1_score))\n",
        "  print(\" CNN Training Precision: {:.4f}\".format(precision))\n",
        "  print(\" CNN Training Recall: {:.4f}\".format(recall))\n",
        "\n",
        "\n",
        "  loss, accuracy, f1_score, precision, recall = model.evaluate(X_test, y_test, verbose=1)\n",
        "  print(\" CNN Test Loss: {:.4f}\".format(loss))\n",
        "  print(\" CNN Test Accuracy: {:.4f}\".format(accuracy))\n",
        "  print(\" CNN Test f1 score: {:.4f}\".format(f1_score))\n",
        "  print(\" CNN Test Precision: {:.4f}\".format(precision))\n",
        "  print(\" CNN Test Recall: {:.4f}\".format(recall))\n",
        "\n",
        "\n",
        "  test_score = model.evaluate(X_test, y_test, verbose=0)\n",
        "  print('Test loss: {}, Test accuracy: {}'.format(test_score[0], test_score[1]))\n",
        "\n",
        "  number_of_epochs_it_ran = len(history.history['accuracy'])\n",
        "\n",
        "  fig, axarr = plt.subplots(figsize=(12,6), ncols=2)\n",
        "  axarr[0].plot(range(1, number_of_epochs_it_ran+1), history.history['accuracy'], label='train score')\n",
        "  axarr[0].plot(range(1, number_of_epochs_it_ran+1), history.history['val_accuracy'], label='test score')\n",
        "  axarr[0].set_xlabel('Number of Epochs', fontsize=18)\n",
        "  axarr[0].set_ylabel('Accuracy', fontsize=18)\n",
        "  axarr[0].set_ylim([0,1])\n",
        "  axarr[1].plot(range(1, number_of_epochs_it_ran+1), history.history['accuracy'], label='train score')\n",
        "  axarr[1].plot(range(1, number_of_epochs_it_ran+1), history.history['val_accuracy'], label='test score')\n",
        "  axarr[1].set_xlabel('Number of Epochs', fontsize=18)\n",
        "  axarr[1].set_ylabel('Accuracy', fontsize=18)\n",
        "  axarr[1].set_ylim([0.80,1])\n",
        "  plt.legend()\n",
        "  plt.show()\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dtFcwfT4-Bh8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#batch,epoch,lr must be added to the function params.\n",
        "\n",
        "def indexing (arr):\n",
        "  new=[]\n",
        "  for a in  arr:\n",
        "    new.append(a-1)\n",
        "\n",
        "  return np.array(new)\n",
        "\n",
        "y_train=indexing(y_train)\n",
        "y_test=indexing(y_test)\n",
        "\n",
        "def one_hot(y_, n_classes=6):\n",
        "    # Function to encode neural one-hot output labels from number indexes\n",
        "    # e.g.:\n",
        "    # one_hot(y_=[[5], [0], [3]], n_classes=6):\n",
        "    #     return [[0, 0, 0, 0, 0, 1], [1, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0, 0]]\n",
        "\n",
        "    y_ = y_.reshape(len(y_))\n",
        "    return np.eye(n_classes)[np.array(y_, dtype=np.int32)]  # Returns FLOATS\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tx64_Khu6esm",
        "colab_type": "code",
        "outputId": "12acebe2-a1c5-4e71-ede1-e4f1d03ec6b3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "\n",
        "print(x_train.shape,y_train.shape,x_test.shape,y_test.shape)\n",
        "#history_lstm=lstm_train(x_train,one_hot(y_train),x_test,one_hot(y_test))\n",
        "print(one_hot(y_test))\n",
        "\n",
        "history_lstm=lstm_train(x_train,one_hot(y_train),x_test,one_hot(y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(7352, 128, 9) (7352,) (2947, 128, 9) (2947,)\n",
            "[[1. 0. 0. 0. 0. 0.]\n",
            " [0. 0. 0. 0. 0. 1.]\n",
            " [0. 0. 1. 0. 0. 0.]\n",
            " ...\n",
            " [0. 0. 0. 1. 0. 0.]\n",
            " [0. 0. 0. 0. 0. 1.]\n",
            " [0. 0. 1. 0. 0. 0.]]\n",
            "(None, 128, 9)\n",
            "(None, 64)\n",
            "(None, 128)\n",
            "(None, 6)\n",
            "Train on 6616 samples, validate on 736 samples\n",
            "Epoch 1/50\n",
            "6616/6616 [==============================] - 30s 5ms/step - loss: 0.7793 - accuracy: 0.6601 - f1_m: 0.6135 - precision_m: 0.7958 - recall_m: 0.5323 - val_loss: 0.3817 - val_accuracy: 0.8247 - val_f1_m: 0.8296 - val_precision_m: 0.8492 - val_recall_m: 0.8112\n",
            "Epoch 2/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.3947 - accuracy: 0.8407 - f1_m: 0.8402 - precision_m: 0.8576 - recall_m: 0.8249 - val_loss: 0.1488 - val_accuracy: 0.9443 - val_f1_m: 0.9440 - val_precision_m: 0.9440 - val_recall_m: 0.9440\n",
            "Epoch 3/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.2367 - accuracy: 0.9167 - f1_m: 0.9167 - precision_m: 0.9198 - recall_m: 0.9137 - val_loss: 0.1447 - val_accuracy: 0.9524 - val_f1_m: 0.9518 - val_precision_m: 0.9518 - val_recall_m: 0.9518\n",
            "Epoch 4/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.3816 - accuracy: 0.8641 - f1_m: 0.8612 - precision_m: 0.8791 - recall_m: 0.8456 - val_loss: 0.2071 - val_accuracy: 0.9130 - val_f1_m: 0.9153 - val_precision_m: 0.9165 - val_recall_m: 0.9141\n",
            "Epoch 5/50\n",
            "6616/6616 [==============================] - 29s 4ms/step - loss: 0.2070 - accuracy: 0.9247 - f1_m: 0.9246 - precision_m: 0.9282 - recall_m: 0.9212 - val_loss: 0.1655 - val_accuracy: 0.9402 - val_f1_m: 0.9419 - val_precision_m: 0.9438 - val_recall_m: 0.9401\n",
            "Epoch 6/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1495 - accuracy: 0.9400 - f1_m: 0.9399 - precision_m: 0.9413 - recall_m: 0.9385 - val_loss: 0.1261 - val_accuracy: 0.9538 - val_f1_m: 0.9550 - val_precision_m: 0.9569 - val_recall_m: 0.9531\n",
            "Epoch 7/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1397 - accuracy: 0.9454 - f1_m: 0.9457 - precision_m: 0.9469 - recall_m: 0.9444 - val_loss: 0.1248 - val_accuracy: 0.9416 - val_f1_m: 0.9427 - val_precision_m: 0.9427 - val_recall_m: 0.9427\n",
            "Epoch 8/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1246 - accuracy: 0.9513 - f1_m: 0.9514 - precision_m: 0.9520 - recall_m: 0.9509 - val_loss: 0.1209 - val_accuracy: 0.9552 - val_f1_m: 0.9557 - val_precision_m: 0.9557 - val_recall_m: 0.9557\n",
            "Epoch 9/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1107 - accuracy: 0.9530 - f1_m: 0.9530 - precision_m: 0.9532 - recall_m: 0.9529 - val_loss: 0.1164 - val_accuracy: 0.9484 - val_f1_m: 0.9479 - val_precision_m: 0.9479 - val_recall_m: 0.9479\n",
            "Epoch 10/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1215 - accuracy: 0.9519 - f1_m: 0.9519 - precision_m: 0.9522 - recall_m: 0.9515 - val_loss: 0.1352 - val_accuracy: 0.9429 - val_f1_m: 0.9446 - val_precision_m: 0.9453 - val_recall_m: 0.9440\n",
            "Epoch 11/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1130 - accuracy: 0.9510 - f1_m: 0.9510 - precision_m: 0.9512 - recall_m: 0.9508 - val_loss: 0.1203 - val_accuracy: 0.9620 - val_f1_m: 0.9622 - val_precision_m: 0.9622 - val_recall_m: 0.9622\n",
            "Epoch 12/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1423 - accuracy: 0.9450 - f1_m: 0.9455 - precision_m: 0.9462 - recall_m: 0.9449 - val_loss: 0.1340 - val_accuracy: 0.9538 - val_f1_m: 0.9549 - val_precision_m: 0.9581 - val_recall_m: 0.9518\n",
            "Epoch 13/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1392 - accuracy: 0.9472 - f1_m: 0.9475 - precision_m: 0.9486 - recall_m: 0.9464 - val_loss: 0.1201 - val_accuracy: 0.9511 - val_f1_m: 0.9505 - val_precision_m: 0.9505 - val_recall_m: 0.9505\n",
            "Epoch 14/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1218 - accuracy: 0.9510 - f1_m: 0.9513 - precision_m: 0.9517 - recall_m: 0.9509 - val_loss: 0.1126 - val_accuracy: 0.9524 - val_f1_m: 0.9524 - val_precision_m: 0.9531 - val_recall_m: 0.9518\n",
            "Epoch 15/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1135 - accuracy: 0.9531 - f1_m: 0.9536 - precision_m: 0.9538 - recall_m: 0.9533 - val_loss: 0.1077 - val_accuracy: 0.9579 - val_f1_m: 0.9583 - val_precision_m: 0.9583 - val_recall_m: 0.9583\n",
            "Epoch 16/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1038 - accuracy: 0.9587 - f1_m: 0.9587 - precision_m: 0.9587 - recall_m: 0.9587 - val_loss: 0.1138 - val_accuracy: 0.9497 - val_f1_m: 0.9485 - val_precision_m: 0.9492 - val_recall_m: 0.9479\n",
            "Epoch 17/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1025 - accuracy: 0.9575 - f1_m: 0.9574 - precision_m: 0.9574 - recall_m: 0.9573 - val_loss: 0.0993 - val_accuracy: 0.9606 - val_f1_m: 0.9609 - val_precision_m: 0.9609 - val_recall_m: 0.9609\n",
            "Epoch 18/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1011 - accuracy: 0.9575 - f1_m: 0.9574 - precision_m: 0.9576 - recall_m: 0.9573 - val_loss: 0.0942 - val_accuracy: 0.9606 - val_f1_m: 0.9609 - val_precision_m: 0.9609 - val_recall_m: 0.9609\n",
            "Epoch 19/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.0967 - accuracy: 0.9593 - f1_m: 0.9591 - precision_m: 0.9592 - recall_m: 0.9589 - val_loss: 0.1001 - val_accuracy: 0.9538 - val_f1_m: 0.9544 - val_precision_m: 0.9544 - val_recall_m: 0.9544\n",
            "Epoch 20/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1005 - accuracy: 0.9559 - f1_m: 0.9553 - precision_m: 0.9554 - recall_m: 0.9552 - val_loss: 0.0982 - val_accuracy: 0.9538 - val_f1_m: 0.9518 - val_precision_m: 0.9518 - val_recall_m: 0.9518\n",
            "Epoch 21/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.0976 - accuracy: 0.9548 - f1_m: 0.9548 - precision_m: 0.9550 - recall_m: 0.9547 - val_loss: 0.1181 - val_accuracy: 0.9457 - val_f1_m: 0.9466 - val_precision_m: 0.9466 - val_recall_m: 0.9466\n",
            "Epoch 22/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1043 - accuracy: 0.9575 - f1_m: 0.9572 - precision_m: 0.9573 - recall_m: 0.9570 - val_loss: 0.1022 - val_accuracy: 0.9552 - val_f1_m: 0.9564 - val_precision_m: 0.9570 - val_recall_m: 0.9557\n",
            "Epoch 23/50\n",
            "6616/6616 [==============================] - 28s 4ms/step - loss: 0.1045 - accuracy: 0.9548 - f1_m: 0.9551 - precision_m: 0.9552 - recall_m: 0.9549 - val_loss: 0.1075 - val_accuracy: 0.9484 - val_f1_m: 0.9492 - val_precision_m: 0.9505 - val_recall_m: 0.9479\n",
            "Epoch 24/50\n",
            "6616/6616 [==============================] - 27s 4ms/step - loss: 0.1052 - accuracy: 0.9530 - f1_m: 0.9524 - precision_m: 0.9525 - recall_m: 0.9524 - val_loss: 0.1040 - val_accuracy: 0.9606 - val_f1_m: 0.9616 - val_precision_m: 0.9622 - val_recall_m: 0.9609\n",
            "Epoch 25/50\n",
            "6616/6616 [==============================] - 26s 4ms/step - loss: 0.1008 - accuracy: 0.9560 - f1_m: 0.9558 - precision_m: 0.9558 - recall_m: 0.9558 - val_loss: 0.1073 - val_accuracy: 0.9538 - val_f1_m: 0.9531 - val_precision_m: 0.9531 - val_recall_m: 0.9531\n",
            "Restoring model weights from the end of the best epoch\n",
            "Epoch 00025: early stopping\n",
            "7352/7352 [==============================] - 7s 940us/step\n",
            " CNN Training Loss: 0.0957\n",
            " CNN Training Accuracy: 0.9581\n",
            " CNN Training f1 score: 0.9580\n",
            " CNN Training Precision: 0.9581\n",
            " CNN Training Recall: 0.9580\n",
            "2947/2947 [==============================] - 3s 927us/step\n",
            " CNN Test Loss: 0.3476\n",
            " CNN Test Accuracy: 0.8989\n",
            " CNN Test f1 score: 0.8992\n",
            " CNN Test Precision: 0.9007\n",
            " CNN Test Recall: 0.8978\n",
            "Test loss: 0.3475944659782338, Test accuracy: 0.8988802433013916\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QhINuFHqfiKV",
        "colab_type": "text"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ha4GqtHVE-nv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}