Table 3. Results of FS Methods for SC-1 and SC-2.

Feature selection	Pre process	Classifier	Wrapper	Number of features	AUPRC	AUROC
SC-1 Ph.1
ReliefF*	STPE	LR	KNN	55	0.9341	0.8235
ReliefF	AF	LR	XGB	10	0.9264	0.7745
SC-1 Ph.3
Fisher Score*	STPE	KNN	XGB	60	0.9746	0.9167
F Score*	STPE	KNN	XGB	60	0.9746	0.9167
mRMR*	STPE	SVM	XGB	275	0.9706	0.9118
mRMR*	STPE	KNN	KNN	6,805	0.9632	0.8775
Fisher Score*	STPE	RF	LR	20,481	0.9628	0.8725
Gini Index*	STPE	KNN	KNN	12,302	0.9572	0.8627
ReliefF*	STPE	XGB	KNN	18,913	0.9502	0.8725
ReliefF	STPE	LR	LR	22,277	0.9498	0.8627
Fisher Score*	AF	KNN	XGB	40	0.9429	0.8235
mRMR	AF	SVM	KNN	16	0.9325	0.7745
SC-2 Ph.1
Fisher Score*	AF	KNN	KNN	17,566	0.8515	0.8712
F Score*	AF	KNN	KNN	17,566	0.8515	0.8712
Gini Index*	AF	LR	KNN	14,673	0.8365	0.8561
Chi Square*	AF	XGB	LR	8	0.8187	0.7765
Chi Square	STPE	KNN	XGB	54	0.8112	0.7689
Chi Square	AF	LR	KNN	5	0.8039	0.7879
SC-2 Ph.3
Fisher Score*	AF	KNN	KNN	18,084	0.8956	0.8561
F Score*	AF	KNN	KNN	18,084	0.8956	0.8561
Gini Index*	AF	SVM	LR	116	0.8908	0.8939
Chi Square	AF	LR	LR	19	0.8759	0.8712
Chi Square	AF	LR	KNN	6	0.8675	0,8561
Chi Square	STPE	KNN	XGB	180	0.8595	0.8333
ReliefF	STPE	KNN	LR	12,206	0.8518	0.8447
Fisher Score*	STPE	KNN	KNN	22,214	0.8497	0.8598
Gini Index*	AF	LR	KNN	8,495	0.8462	0.8333
Gini Index*	AF	SVM	XGB	4	0.8428	0.8258
ReliefF	AF	LR	XGB	92	0.821	0.8106

Note:

After the features are ranked by a filtering approach, a wrapper algorithm is utilized to select the best feature subset. Wrapper column indicates the prediction algorithm used in wrapper method. Number of Features column represents the number of distinct features selected. An asterisk (*) indicates that the hyper-parameters were not optimized.