A Hybrid Method Based on Extreme Learning Machine and Self Organizing Map for Pattern Classification

Table 2

Results of Experiments on classification problem.

Datasets	Algorithms	Training time (s)	Testing accuracy	Hidden nodes
Ionosphere	BP	33.944864	82.8571	20
	ELM	0.667364	87.6190	40
	TROP-ELM	0.6517	89.2900	51
	Proposed	0.396109	92.3810	31
Iris	BP	29.232330	93.3333	15
	ELM	0.235005	95.5556	40
	TROP-ELM	0.0738	96.6700	59
	Proposed	0.052971	100	15
Wine	BP	31.336370	94.2308	18
	ELM	0.218872	96.1538	35
	TROP-ELM	0.1242	96.5800	84
	Proposed	0.092291	100	22
Balance	BP	136.802525	76.1905	10
	ELM	1.121717	83.0688	28
	TROP-ELM	0.3224	87.2100	56
	Proposed	0.474296	87.5661	13
Zoo	BP	34.132912	80.6452	10
	ELM	0.082092	93.5484	15
	TROP-ELM	0.0316	94.5000	18
	Proposed	0.030456	97.2350	7
Image segmentation	BP	428.141558	87.8582	15
	ELM	24.629949	91.3008	90
	TROP-ELM	207.8026	90.4300	187
	Proposed	11.964243	96.4131	70
Ecoli	BP	55.678908	71.4286	15
	ELM	0.646599	85.9890	40
	TROP-ELM	0.5869	92.0700	90
	Proposed	0.252657	93.6813	26
Multiple features	BP	226.625993	90.9000	12
	ELM	67.085971	97.6833	180
	TROP-ELM	190.431	98.4300	338
	Proposed	40.753813	99.0333	121
Jaffe	BP	427.894980	75.4464	20
	ELM	1.778349	81.9196	130
	TROP-ELM	1.668231	83.4550	240
	Proposed	1.530555	84.3750	99