Table 2 The parameters of the forecasting models

SARIMA

Seasonal period

Preset

5

AR

Partial autocorrelation function

[0,5]

I

Augmented dickeye-fuller test

0 or 1

Ma

Autocorrelation function

[0,5]

SVR

Regularization coefficient c

Grid search

[1,300]

Kernel parameter g

Grid search

[\(2^{-5}\), \(2^{5}\)]

ELM& BPNN

Input dimension

Preset

–

Number of hidden layer nodes

Trial and error approach

24

Output dimension

Preset

1

Maximum of epochs

Preset

100

Number of parameters (BPNN)

–

361

LSTM

Input dimension

Preset

–

Number of hidden layer nodes

Trial and error approach

24

Output dimension

Preset

1

Maximum of epochs

Preset

100

Number of parameters

–

3673

TCN

Input dimension

Preset

–

Nb_filters

Trial and error approach

32

Kernel_size

Trial and error approach

2

Output dimension

Preset

1

Maximum of epochs

Preset

100

Number of parameters

–

24225