Table 3 Hyperparameters of BP neural network optimized by optuna.

Number of hidden layers

[1, 2]

2

Number of neurons in the first hidden layer

[50, 200]

175

Number of neurons in the second hidden layer

[50, 200]

124

Activation function

[‘relu’, ‘tanh’]

‘relu’

L2 regularization coefficient

[1E-6, 1E-3]

2.285E-5

Initial learning rate

[1E-5, 1E-1]

1.075E-3

Learning rate strategy

[‘constant’, ‘adaptive’]

constant

Maximum iteration number

[2000, 5000]

2656