Table 2.

Performance of feature selection.

			RF
Strategy	Start up	# peaks	Accuracy	AUC	p-value
Stepwise	AUC	21	0.918 ± 0.024	0.921 ± 0.025	0.0079
	Kendall's tau	26	0.906 ± 0.008	0.897 ± 0.049	0.2222
	KW	20	0.902 ± 0.030	0.917 ± 0.042	0.0952
	IMP	27	0.910 ± 0.008	0.926 ± 0.020	0.0079
Forward	AUC	18	0.897 ± 0.032	0.896 ± 0.026	0.0952
	Kendall's tau	35	0.901 ± 0.024	0.893 ± 0.047	0.2222
	FE	25	0.902 ± 0.023	0.898 ± 0.035	0.0556
	KW	26	0.906 ± 0.031	0.902 ± 0.061	0.2222
	IMP-ACC	22	0.882 ± 0.032	0.864 ± 0.059	0.7533
	IMP-GINI	28	0.874 ± 0.034	0.836 ± 0.037	0.5476
No		583	0.866 ± 0.014	0.848 ± 0.037	-

Mean ± standard deviation accuracy and AUC of the 5-fold cross validation of RF using different numbers of peaks selected by the forward and stepwise feature selection strategies using different orders of peaks and the corresponding performance using RF. AUC, area under the curve; FE, Fisher's exact test; KW, Kruskal-Wallis test; IMP-ACC, importance measure calculated by mean decreased accuracy using RF; IMP-GINI, importance measure calculated by mean decreased impurity using RF.