Table 2.

Evaluation metrics of seven feature selection approaches and the full feature set with a linear support vector machine (SVM) classifier.

Method	Accuracy	Sensitivity	Specificity	AUC	Feature number	p
Full feature set	53.97%	56.14%	55.07%	0.5362	46	0.226
F score	55.56%	70.17%	47.83%	0.5751	11	0.062
T-test	56.35%	78.95%	37.68%	0.5238	11	0.329
Gini Index	57.14%	64.91%	52.17%	0.5700	13	0.088
SFS	74.60%	75.44%	75.36%	0.7442	5	<0.001
SFFS	71.43%	80.70%	66.67%	0.7732	6	<0.001
SBFS	58.73%	82.46%	46.38%	0.6293	30	0.007
SBE	81.75%	84.21%	81.15%	0.8241	17	<0.001