Table 1. Comparison of the classification performance achieved with and without feature selection, for (a) the DLBCL dataset, (b) the Glioma dataset and (c) the Uterus dataset.

The comparison is carried out for each of the considered selection methods (by retaining 2% of the original features). The values in bold turned out to be significantly different from the baseline according to the Wilcoxon signed-rank test with Holm-Bonferroni correction (p-values in brackets).

(a)	Baseline	CORR	IG	GR	RF	SVM-AW	SVM-RFE
MCC	0.50	0.72	0.71	0.71	0.80	0.85	0.80
		(0.0024)	(0.0085)	(0.0035)	(0.0006)	(0.0008)	(0.0008)
G-mean	0.55	0.82	0.79	0.78	0.86	0.87	0.84
		(0.0010)	(0.0017)	(0.0010)	(0.0006)	(0.0007)	(0.0008)
F-measure	0.5	0.76	0.75	0.73	0.82	0.85	0.81
		(0.0019)	(0.0038)	(0.0023)	(0.0009)	(0.0007)	(0.0008)
(b)	Baseline	CORR	IG	GR	RF	SVM-AW	SVM-RFE
MCC	0.08	0.58	0.58	0.55	0.52	0.30	0.37
		(0.0012)	(0.0012)	(0.0017)	(0.0024)	(0.0267)	(0.0131)
G-mean	0.09	0.59	0.59	0.57	0.52	0.31	0.37
		(0.0012)	(0.0012)	(0.0014)	(0.0024)	(0.0267)	(0.0131)
F-measure	0.08	0.58	0.58	0.57	0.52	0.30	0.37
		(0.0012)	(0.0012)	(0.0014)	(0.0024)	(0.0267)	(0.0131)
(c)	Baseline	CORR	IG	GR	RF	SVM-AW	SVM-RFE
MCC	0.34	0.50	0.48	0.47	0.50	0.46	0.41
		(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(0.0004)
G-mean	0.38	0.63	0.62	0.61	0.63	0.52	0.47
		(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(0.0002)
F-measure	0.26	0.51	0.49	0.48	0.51	0.41	0.35
		(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(<0.0001)	(0.0002)