Sample-Based Extreme Learning Machine with Missing Data

Table 3

Performance comparison of S-ELM and MF-ELM over different missing ratios on the fingerprint data set.

Target class	Missing ratio	S-ELM		MF-ELM
		Accuracy ± std.	Time (Sec.)	Accuracy ± std.	Time (Sec.)
Loop	10%	0.8256 ± 0.0736	89.37	0.7967 ± 0.0985	85.72
	20%	0.8029 ± 0.1253	85.91	0.7265 ± 0.1371	86.28
	30%	0.7694 ± 0.1191	90.43	0.6303 ± 0.1039	87.93
Whorl	10%	0.8137 ± 0.0627	67.28	0.8129 ± 0.0922	60.72
	20%	0.7436 ± 0.0963	71.32	0.7269 ± 0.1167	67.49
	30%	0.7102 ± 0.0849	69.09	0.6359 ± 0.1076	68.25
Arch	10%	0.8361 ± 0.0852	102.7	0.8057 ± 0.0937	94.73
	20%	0.7659 ± 0.0974	98.31	0.7561 ± 0.1170	96.83
	30%	0.7192 ± 0.0827	100.3	0.6574 ± 0.1207	96.67