Table 4.

Description of best-selected parameters of deep learning models.

Algorithm	Number of units	Number of layers	Number of full connected units	Number of fully connected layers	Learning rate	Number of epochs (in each fold)	Batch size	Optimizer	Loss function	Dropout	Activation function of layers	Last layer’s activation function
DNN	–	–	34, 68, 136, 272, 544, 1088, 544, 272, 136, 68, 34, 1	1–12	1e-4	50	21	Binary cross-entropy	Adam	–	Tanh	Sigmoid
CNN	128, 64	1, 2	32, 16, 1	1–3	1e-4	50	21	Binary cross-entropy	Adam	0.1	Tanh	Sigmoid
RNN	64	1	32, 16, 1	1–3	1e-4	50	21	Binary cross-entropy	Adam	0.1	Tanh	Sigmoid
LSTM	64	1	32, 16, 1	1–3	1e-4	50	21	Binary cross-entropy	Adam	0.1	Tanh	Sigmoid