Table 3.

Results for scenario 1, displaying all performance indicators averaged over the 50 repetitions

FSM	AUC	Se	Spe	NSIF	NSNIF	Stability	Runtime (s)	Rank
FULL	85.77	86.11	86.18	3.00	22.00	1.00	0.04	11
CFS	83.76	84.30	84.02	1.82	0.00	0.849	0.23	6
CHI2	85.09	86.09	84.88	1.98	0.00	0.980	0.16	4
GAMMA_BACK	85.92	85.98	86.59	2.92	1.62	0.805	5.68	3
GAMMA_BF	85.92	85.98	86.59	2.92	1.62	0.805	0.45	2
GAMMA_FORW	85.92	85.98	86.59	2.92	1.62	0.805	0.44	1
LASSO	84.02	84.48	84.21	2.90	2.26	0.504	2.38	5
RFI	85.25	86.21	85.05	2.00	0.00	1.00	9.22	7
STEP	85.87	86.41	86.05	3.00	3.16	0.404	1.75	8
SU	77.28	78.18	77.27	1.08	0.00	0.925	0.14	9
SVM-RFE	85.91	86.16	86.42	3.00	2.44	0.489	23.36	10

The best value of each indicator is highlighted with bold and underline text. The second-best value is highlighted in bold. The indicators for the FULL model are not highlighted in the table since the focus of the study is to compare the feature selection methods

Abbreviations: FSM Feature Selection Methods, Se Sensitivity, Spe Specificity, NSIF Number of Selected Informative Features, NSNIF Number of Selected Non-Informative Features