SGB-ELM: An Advanced Stochastic Gradient Boosting-Based Ensemble Scheme for Extreme Learning Machine

Table 4

The comparison results between SGB-ELM and other representative algorithms on 5 classification datasets.

Dataset	Algorithm	Training time	Training accuracy (Dev)	Testing accuracy (Dev)	Hidden nodes	Iterations
Segmentation; ()	ELM	0.0431	0.9465 0.0055	0.9351 0.0064	180	N/A
	V-ELM	0.4487	0.9463 0.0061	0.9374 0.0060	180	7
	EN-ELM	43.9234	0.9472 0.0048	0.9353 0.0063	180 ()	50
	Bagging	3.1853	0.9474 0.0035	0.9353 0.0050	180	50
	Adaboost	3.5372	0.9853 0.0052	0.9466 0.0067	100	Max = 100
	SGB-ELM	134.2969	0.9761 0.0030	0.9558 0.0049	100 ()	100
Texture; ()	ELM	0.0338	0.9954 0.0011	0.9945 0.0016	100	N/A
	V-ELM	0.4275	0.9965 0.0007	0.9950 0.0013	100	7
	EN-ELM	44.4969	0.9963 0.0008	0.9946 0.0011	100 ()	50
	Bagging	3.0959	0.9965 0.0007	0.9957 0.0013	100	50
	Adaboost	10.3628	0.9996 0.0017	0.9972 0.0024	60	Max = 100
	SGB-ELM	193.2019	0.9992 0.0005	0.9982 0.0008	60 ()	100
Spambase; ()	ELM	0.0459	0.9174 0.0044	0.9080 0.0061	150	N/A
	V-ELM	0.4213	0.9192 0.0042	0.9115 0.0053	150	7
	EN-ELM	62.6000	0.9183 0.0054	0.9071 0.0060	150 ()	50
	Bagging	2.9869	0.9219 0.0039	0.9145 0.0051	150	50
	Adaboost	7.6875	0.9620 0.0046	0.9234 0.0072	100	Max = 100
	SGB-ELM	129.0922	0.9522 0.0033	0.9222 0.0043	100 ()	100
Banana; ()	ELM	0.0550	0.6838 0.0253	0.6787 0.0263	180	N/A
	V-ELM	0.4906	0.6860 0.0261	0.6848 0.0264	180	7
	EN-ELM	67.5578	0.6821 0.0227	0.6780 0.0250	180 ()	50
	Bagging	3.3253	0.6808 0.0164	0.6777 0.0178	180	50
	Adaboost	7.5100	0.7457 0.0288	0.7448 0.0320	100	Max = 100
	SGB-ELM	133.0791	0.7610 0.0085	0.7563 0.0082	100 ()	100
Ring; ()	ELM	0.0897	0.9492 0.0024	0.9418 0.0035	200	N/A
	V-ELM	0.7609	0.9532 0.0024	0.9466 0.0032	200	7
	EN-ELM	114.0641	0.9517 0.0028	0.9418 0.0031	200 ()	50
	Bagging	5.5241	0.9539 0.0027	0.9468 0.0030	200	50
	Adaboost	17.2109	0.9940 0.0023	0.9524 0.0038	150	Max = 100
	SGB-ELM	363.7976	0.9750 0.0021	0.9567 0.0027	150 ()	100