Comparative evaluation of set-level techniques in predictive classification of gene expression samples

BMC Bioinformatics

Table 6 Generic feature selection (set-level)

Sets	Methods		Accuracy				Avg Subrank
	Selection	Aggrgt	Median	Avg	σ	Iqr
1:10	SVM-RFE	SVD	88.3	80.6	17.3	33.0	17.6
1:10	IG	SVD	87.0	79.0	18.7	31.6	17.4
1:10	IG	AVG	84.6	78.2	18.6	33.4	18.7
1:10	SVM-RFE	AVG	84.4	79.2	17.1	31.2	19.2
1:10	SVM-RFE	SetSig	82.5	78.7	17.0	31.2	19.4
1	IG	SVD	80.8	76.3	17.7	33.1	22.5
1:10	IG	SetSig	80.0	77.1	17.4	33.2	20.8
1	SVM-RFE	SetSig	71.8	73.7	15.8	26.4	23.3
1	SVM-RFE	SVD	71.5	74.4	17.4	30.3	23.0
1	IG	AVG	70.9	74.0	18.6	33.1	24.1
1	SVM-RFE	AVG	70.8	72.5	15.4	26.6	24.4
1	IG	SetSig	66.2	68.8	16.2	25.0	28.9

Performance of the set level classification strategy using the information gain and SVM-RFE heuristics for ranking gene sets. Column descriptions are as in Table 3.

ISSN: 1471-2105