Table 2.

The classification performance (in percentage) of four machine learning methods on GM.

			GM
	RetainedFeas	OptimalFeas	Sen	Spec	Acc	AUC
SPEC	9,396	8,045	86.93 ± 2.41	91.36 ± 2.94	88.61 ± 1.86	89.15 ± 2.53
ReliefF	9,396	7,809	87.29 ± 3.54	92.24± 3.69	89.58 ± 2.39	89.77 ± 3.58
RFE	5,285	2,841	92.42 ± 2.62	85.85 ± 2.46	88.56 ± 1.99	89.14 ± 2.55
STABLASSO	3,547	2368	90.41 ± 3.21	79.55 ± 2.39	84.98 ± 1.99	84.98 ± 2.98

^*RetainedFeas, The number of features retained by different feature selection methods; OptimalFeas, the optimal feature subsets selected from retained features; Sen, sensitivity; Spec, specificity; Acc, accuracy; AUC, The area under ROC curve. The best performance for each indicator is shown in bold.