BMC Bioinformatics

Table 11 Average performance measures on the Van't Veer et al. [23] (breast cancer) data set after removing samples suspected for measurement noise.

From: Outcome prediction based on microarray analysis: a critical perspective on methods

	Filter Method	Wrapper Methods					Hybrid Methods
Criterion	Fisher's Ratio	RFE LNW-GD	RFE-SVM	RFE-LSSVM	RFE-RR	RFE-FLDA	REF-LNW1	REF-LNW2	RFE-FSVs-7DK
Number of Genes	61	22	32	45	7	28	44	64	73
Per Subject Acc(P)-CV	0.80	0.71	0.69	0.76	0.59	0.65	0.74	0.88	0.78
Per Subject Acc(P)-Test	0.68	0.74	0.62	0.68	0.68	0.64	0.77	0.69	0.76
Per Run Acc(R)-CV	0.74	0.63	0.62	0.65	0.55	0.576	0.69	0.74	0.70
Per Run Test Acc(R)-Test	0.68	0.74	0.62	0.68	0.68	0.641	0.77	0.69	0.76
Per Run All Acc(R)-Overall	0.69	0.71	0.62	0.67	0.65	0.625	0.74	0.70	0.74

All measures are significantly increased compared to those of Table 7 with the removal of only a few samples.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com