Table 1.

The performance among GEDFN + SVM, RF + SVM, GEDFN + DF and RF + DF.

#	GEDFN + SVM			RF + SVM			GEDGN+DF			RF+DF
	P	R	F1	P	R	F1	P	R	F1	P	R	F1
10	0.733	1	0.846	0.675	1	0.806	0.733	1	0.846	0.785	0.871	0.825
15	0.745	1	0.854	0.675	1	0.806	0.745	1	0.854	0.722	0.909	0.800
20	0.752	1	0.858	0.675	1	0.806	0.750	0.991	0.854	0.717	0.927	0.805
25	0.706	1	0.828	0.675	1	0.806	0.705	0.991	0.824	0.765	0.907	0.829
30	0.707	1	0.828	0.675	1	0.806	0.707	0.983	0.823	0.718	0.957	0.821
35	0.698	1	0.822	0.675	1	0.806	0.698	1	0.822	0.692	0.977	0.810
40	0.704	1	0.826	0.675	1	0.806	0.709	0.985	0.824	0.706	0.962	0.813
45	0.707	1	0.828	0.675	1	0.806	0.707	1	0.828	0.687	0.991	0.811
50	0.697	1	0.822	0.675	1	0.806	0.697	1	0.822	0.695	0.974	0.810

#, number of top features; P, precision; R, recall; F1= $\frac{2\times P\times R}{P+R}$ . The best F1 scores are marked as bold.