Soft Sensor Modeling Method Based on SPA-GWO-SVR for Marine Protease Fermentation Process

Table 3

Comparison of different variable selection methods.

Key variables	Variable selection method	Training set		Test set
		R^2	RMSE	R^2	RMSE
X	Full	0.9623	2.4120	0.9672	2.5362
	GRA	0.9742	1.9665	0.9726	1.9965
	SPA	0.9851	1.9057	0.9832	1.9430
S	Full	0.9664	1.8402	0.9623	1.8635
	GRA	0.9690	1.1521	0.9687	1.1854
	SPA	0.9798	1.1422	0.9772	1.1693
P	Full	0.9682	0.1701	0.9678	0.1726
	GRA	0.9734	0.0852	0.9705	0.0921
	SPA	0.9845	0.0389	0.9828	0.0401