9_cnn_rnn_5_with_f1.py

#*****************************************************************************
# Copyright 2019 yoga suhas
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#******************************************************************************

import os
import glob
import pandas as pd
import numpy as np
from functools import partial
from keras.models import Sequential
from keras.layers import BatchNormalization
from keras.layers import Flatten, LSTM, Conv1D, MaxPooling1D, Dropout, Dense
from keras.utils import to_categorical
from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, classification_report

length = 2000
num_epoch = 10
batch_size = 32
index = iter(range(5))
NTC_LABELS = {}

def pad_normalize(s):
    """Collect 1000 bytes from packet payload. If payload length is less than
    1000 bytes, pad zeroes at the end. Then convert to integers and normalize."""

    if len(s) < length:
        s += '00' * (length-len(s))
    else:
        s = s[:length]
    return [float(int(s[i]+s[i+1], 16)/255) for i in range(0, length, 2)]


def fetch_file(file_name, label):
    df = pd.read_csv(file_name, index_col=None, header=0)
    df.columns = ['data']
    df['label'] = label
    return df


def preprocess(path):
    files = glob.glob(os.path.join(path, '*.txt'))
    list_ = []
    for f in files:
        label = f.split('/')[-1].split('.')[0]
        NTC_LABELS[label] = next(index)
        labelled_df = partial(fetch_file, label=NTC_LABELS[label])
        list_.append(labelled_df(f))
    df = pd.concat(list_, ignore_index=True)
    return df


def build_model(X_train, y_train, X_test, y_test, num_epoch, batch_size):
    activation = 'relu'
    num_classes = len(NTC_LABELS)
    model = Sequential()
    model.add(Conv1D(32, strides=1, input_shape=(1000, 1), activation=activation, kernel_size=3, padding='same'))
    model.add(BatchNormalization())
    model.add(Conv1D(32, strides=1, activation=activation, kernel_size=3, padding='same'))
    model.add(MaxPooling1D(2, strides=2, padding='same'))
    model.add(BatchNormalization())
    model.add(Conv1D(64, strides=1, activation=activation, kernel_size=3, padding='same'))
    model.add(BatchNormalization())
    model.add(Conv1D(64, strides=1, activation=activation, kernel_size=3, padding='same'))
    model.add(MaxPooling1D(2, strides=2, padding='same'))
    model.add(Conv1D(128, strides=1, activation=activation, kernel_size=3, padding='same'))
    model.add(Conv1D(64, strides=1, activation=activation, kernel_size=3, padding='same'))
    model.add(LSTM(100,return_sequences=True))
    model.add(Dropout(0.1))
    model.add(LSTM(150,return_sequences=True))
    model.add(Dropout(0.3))
    model.add(Dense(150, activation=activation))
    model.add(Dropout(0.5))
    model.add(Dense(108, activation=activation))
    model.add(Flatten())
    model.add(Dense(num_classes, activation='softmax'))
    model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    print(model.summary())
    model.fit(X_train, y_train, verbose=1, epochs=num_epoch, batch_size=batch_size, validation_data=(X_test, y_test))
    return model


def main():
    df = preprocess(path='Dataset')
    df['data'] = df['data'].apply(pad_normalize)
    X_train, X_test, y_train, y_test = train_test_split(df['data'], df['label'],
                                                        test_size=0.3, random_state=4)
    X_train = X_train.apply(pd.Series)
    X_test = X_test.apply(pd.Series)
    X_train = X_train.values.reshape(X_train.shape[0], X_train.shape[1], 1)
    X_test = X_test.values.reshape(X_test.shape[0], X_test.shape[1], 1)

    
    model = build_model(X_train, y_train, X_test, y_test, num_epoch, batch_size)

    # predict crisp classes for test set
    y_pred_t = model.predict(X_test, batch_size=32, verbose=1)
    y_pred_bool = np.argmax(y_pred_t, axis=1)
    y_pred = model.predict_classes(X_test)


    # accuracy: (tp + tn) / (p + n)
    accuracy = accuracy_score(y_test, y_pred)
    print('Accuracy: %f' % accuracy)
    # precision tp / (tp + fp)
    precision = precision_score(y_test, y_pred, average='weighted',labels=np.unique(y_pred))
    print('Precision: %f' % precision)
    # recall: tp / (tp + fn)
    recall = recall_score(y_test, y_pred,average='weighted')
    print('Recall: %f' % recall)
    # f1: 2 tp / (2 tp + fp + fn)
    f1 = f1_score(y_test, y_pred, average='weighted',labels=np.unique(y_pred))
    print('F1 score: %f' % f1)
    
    print(classification_report(y_test, y_pred_bool))


if __name__ == '__main__':
    main()