An Incremental Learning Ensemble Strategy for Industrial Process Soft Sensors

Table 3

The performance of the soft sensor model based on different methods with 10-fold cross validation.

Method	datasets			RMSE	ARE	Max ARE
	training	updating	testing
ELM	~	S11	S10	0.7987	0.0698	0.2265
AdaBoost.R				0.5915	0.0313	0.1447
ILES				0.3704	0.0213	0.0689
ELM	~	S11	S9	0.7386	0.0868	0.3773
AdaBoost.R				0.4740	0.0394	0.1846
ILES				0.3309	0.0182	0.0542
ELM	~~	S11	S8	0.8405	0.0772	0.2962
AdaBoost.R				0.5099	0.0295	0.1873
ILES				0.3323	0.0202	0.0572
ELM	~~	S11	S7	0.936	0.0655	0.4255
AdaBoost.R				0.4835	0.219	0.1773
ILES				0.3556	0.0204	0.0525
ELM	~~	S11	S6	0.7342	0.0618	0.3451
AdaBoost.R				0.5198	0.0324	0.1816
ILES				0.3965	0.0221	0.0641
ELM	S1~S4,S6~S10	S11	S5	0.8533	0.0872	0.3546
AdaBoost.R				0.4672	0.0336	0.1724
ILES				0.3531	0.0195	0.0499
ELM	~~	S11	S4	0.7752	0.0749	0.3249
AdaBoost.R				0.4105	0.0362	0.1924
ILES				0.3934	0.0209	0.0748
ELM	~~	S11	S3	0.8430	0.0823	0.3281
AdaBoost.R				0.5773	0.0525	0.1827
ILES				0.3656	0.0203	0.0610
ELM	~	S11	S2	0.9127	0.1104	0.4302
AdaBoost.R				0.5695	0.0580	0.1891
ILES				0.3625	0.0211	0.0545
ELM	~	S11	S1	0.7905	0.0910	0.2245
AdaBoost.R				0.5045	0.0436	0.1773
ILES				0.3145	0.0191	0.0756