A Novel Reformed Reduced Kernel Extreme Learning Machine with RELIEF-F for Classification

Table 5

The performance between relief-F and RKELM with two sample selection methods.

Data	Model	Accuracy (%)	SD	Time	Sensitivity	Specificity	Precision
German	K-RKELM	83.53	0.01	0.0011	0.5531	0.5881	0.6045
	C-RKELM	81.67	0.02	0.0419	0.9157	0.1538	0.8787
	R-RKELM	85.44	0.01	0.0056	0.6753	0.3940	0.7184
Image	K-RKELM	87.43	0.01	0.0063	0.8694	0.8736	0.8816
	C-RKELM	87.70	0.00	0.2610	0.9634	0.7819	0.8294
	R-RKELM	87.75	0.00	0.2536	0.9040	0.8424	0.8700
Ringnorm	K-RKELM	55.32	0.00	0.1265	0.5556	0.5541	0.7634
	C-RKELM	67.66	0.00	10.2167	1.0000	0.3555	0.6064
	R-RKELM	69.37	0.00	6.1088	0.7986	0.5910	0.7454
Twonorm	K-RKELM	94.52	0.00	0.1482	0.9382	0.9393	0.9195
	C-RKELM	95.23	0.00	9.1533	0.9385	0.9566	0.8735
	R-RKELM	95.35	0.00	5.7655	0.9479	0.9539	0.9096
Waveform	K-RKELM	85.51	0.01	0.0305	0.8454	0.8258	0.8298
	C-RKELM	85.47	0.00	3.7682	0.8757	0.8117	0.9045
	R-RKELM	85.84	0.00	1.6029	0.8594	0.8344	0.8597
HAPT	K-RKELM	89.50	0.09	0.3997	0.9709	0.8914	0.7367
	C-RKELM	90.58	0.06	18.8531	0.8932	0.8309	0.7481
	R-RKELM	92.87	0.06	24.1748	0.9760	0.9197	0.7869
HARUS	K-RKELM	89.79	0.08	0.3253	0.9618	0.8945	0.7552
	C-RKELM	88.83	0.00	20.1623	0.8468	0.8898	0.6087
	R-RKELM	92.81	0.05	20.1696	0.9649	0.9221	0.8029
Smartphone	K-RKELM	86.68	0.06	0.2197	0.8055	0.7888	0.7092
	C-RKELM	86.68	0.00	5.6465	0.7945	0.8057	0.7178
	R-RKELM	86.92	0.06	1.3800	0.8126	0.8090	0.7268