Research Article

Recursive Feature Selection with Significant Variables of Support Vectors

Table 1

Comparison of selection methods on simulation data 1.

Num. of genesRSVMSVMRFESVM-
Test (%)Rec (%)nSVTest (%)Rec (%)nSVTest (%)Rec (%)nSV

600
500
400
300
200
150
100
90
80
70
60
50

Num. of genes: number of selected features in each recursive step. Results of 1000 features are the same and we do not list them here.
Test: test error rate.
Rec: the percentage of recovery informative genes (among 300 informative genes).
nSV: the average number of support vectors selected in each recursive step.