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

Table 4

The performances of Reduced-KELM and Relief-F.

Data	Model	Accuracy (%)	Difference	SD	Time	Sensitivity	Specificity	Precision
German	Reduced-KELM	71.53	8.74%	0.06	0.0013	0.5790	0.4962	0.5175
	Relief-F	80.27		0.03	0.0012	0.5164	0.5108	0.5068
Image	Reduced-KELM	85.70	0.16%	0.03	0.0086	0.8623	0.8510	0.8555
	Relief-F	85.86		0.02	0.0063	0.8340	0.8780	0.8815
Ringnorm	Reduced-KELM	60.06	0.85%	0.02	0.1128	0.5961	0.6079	0.7786
	Relief-F	60.91		0.01	0.1031	0.6135	0.6075	0.7786
Twonorm	Reduced-KELM	94.10	-1.65%	0.01	0.1022	0.9401	0.9377	0.9115
	Relief-F	92.45		0.01	0.0934	0.9245	0.9216	0.8853
Waveform	Reduced-KELM	84.29	0.93%	0.01	0.0474	0.8324	0.8090	0.8148
	Relief-F	85.22		0.01	0.0384	0.8169	0.8223	0.8160
HAPT	Reduced-KELM	88.52	0.95%	0.08	0.4155	0.9580	0.9434	0.8731
	Relief-F	89.47		0.07	0.3753	0.9552	0.8936	0.7479
HARUS	Reduced-KELM	84.02	5.66%	0.07	0.3826	0.9470	0.9419	0.8643
	Relief-F	89.68		0.06	0.3589	0.9410	0.8851	0.7623
Smartphone	Reduced-KELM	85.52	1.03%	0.07	0.1261	0.7749	0.7535	0.7280
	Relief-F	86.55		0.07	0.0951	0.7869	0.7926	0.7532