Table 2 Hyperparameters of classifier.

CMN-ShuffleNet model

Number of input layer:1

convolution layer1, filter size:24

Activation:relu

Output layer1

Optimizer: adam

Activation function: softmax

Batch size:64

Loss = ‘categorical_crossentropy’

Epoch = 50

Metrics = [‘accuracy’]

GRU

batch_size = 128

epochs = 50

input layer = 1

Bidirectional GRU uint1 = 128,

activation = ‘softmax’,

Dropout = 0.2

loss = ‘categorical_crossentropy’

LeNet

Conv2D—> (16, (5, 5), activation = ‘relu’, input_shape = input_shape_))

MaxPooling2D((2, 2)))

Conv2D(32, (5, 5), activation = ‘relu’))

MaxPooling2D((2, 2)))

Flatten

Dense layer (120, activation = ‘relu’)

Dense layer (84, activation = ‘relu’

Bi-LSTM

input layer—> 1

Bidirectional LSTM layer1 :- > lstm unit:128

activation = ‘relu’, return_sequences = True

Dropout = 0.5

batch_size = 28

ShuffleNet

Number of input layer:1

convolution layer1, filter size:24

kernel_size = 3, strides = 2, padding = ‘same’, use_bias = False

Activation:relu

AlexNet-SVM¹⁸

input layer:1

Conv2D(filters = 96, kernel_size = (3, 3), strides = (1, 1), activation = ‘relu’, ),

BatchNormalization(),

MaxPool2D(pool_size = (3, 3), strides = (2, 2)),

ATT-DenseNet¹⁶

convolution 1

maxpooling 1

dense block1

transition layer1

SE block

dense block2

transition layer2