An Approach for Demand Forecasting in Steel Industries Using Ensemble Learning

Table 4

Summary of all extensive experiments to select the best performing preprocessing feature selection methods with the number of features and regression models.

ML method (s)	Preprocessing	Algorithm	n_features	Performance (R^2)
SVR	Raw data	N/A	6	0.898
	Processed data	N/A	4	0.919
		Corr	4	0.931
		PCA	4	0.923
		ICA	4	0.923
RFR	Raw data	N/A	6	0.918
	Processed data	N/A	4	0.929
		Corr	4	0.966
		PCA	4	0.967
		ICA	4	0.952
MLP	Raw data	N/A	6	0.845
	Processed data	N/A	4	0.911
		Corr	4	0.957
		PCA	4	0.961
		ICA	4	0.950
ELM	Raw data	N/A	6	0.841
	Processed data	N/A	4	0.849
		Corr	4	0.942
		PCA	4	0.887
		ICA	4	0.909
GBR	Raw data	N/A	6	0.934
	Processed data	N/A	4	0.944
		Corr	4	0.969
		PCA	4	0.949
		ICA	4	0.934
XGBR	Raw data	N/A	6	0.949
	Processed data	N/A	4	0.953
		Corr	4	0.974
		PCA	4	0.956
		ICA	4	0.952

N/A = none. Note: the best approaches were shown in bold type.