Table 4.

The structure and parameters of the CNN and GRU.

Model	Parameter	Value
CNN	Conv1 layer1	filters: 8, kernel size: 3, activation: “relu”
	Max pooling layer1	pool_size: 2, strides: 2
	Conv1 layer2	filters: 16, kernel size: 5, activation: “relu”
	Max pooling layer2	pool_size: 2, strides: 2
	Conv1 layer3	filters: 32, kernel size: 5, activation: “relu”
	Max pooling layer3	pool_size: 2, strides: 2
	Conv1 layer4	filters: 64, kernel size: 3, activation: “relu”
	Max pooling layer4	pool_size: 2, strides: 2
	Flatten	–
	Dense1	units: 128, activation: “relu”
	Dense2	units: 10, activation: “softmax”
	Loss function	Categorical Cross-entropy
	Optimizer	Adam
	learning rate	0.02
	Epoch	20
	Batch size	256
GRU	GRU layer1	units: 128, return_sequences = True
	GRU layer2	units: 256, return_sequences = False
	Flatten	–
	Dense1	units: 256, activation: “relu”
	Dense2	units: 10, activation: “softmax”
	Loss function	Categorical Cross-entropy
	Optimizer	Adam
	learning rate	0.02
	Epoch	20
	Batch size	256