Table 6.

Performance of reduced models with wrapper feature selection (RFE) and embedded feature selection (RFI)

Performance			Number of features
			15	14	13	12	11
Wrapper (RFE)	Testing accuracy* (%)	RF	77.00	77.03	76.72	76.00	76.29
		ANN	73.78	73.00	72.29	72.93	73.21
		SVM	73.57	73.86	73.50	73.93	73.86
		LR	71.93	71.93	71.46	71.86	72.22
		KNN	69.43	70.14	69.28	72.86	73.64
		DT	67.50	67.57	68.21	69.00	68.43
	F1 score*	RF	0.7062	0.6991	0.7108	0.6949	0.7054
		ANN	0.6609	0.6512	0.6395	0.6447	0.6438
		SVM	0.6627	0.6711	0.6674	0.6708	0.6646
		LR	0.6203	0.6288	0.6207	0.6137	0.6182
		KNN	0.6254	0.6361	0.6224	0.6795	0.6840
		DT	0.6009	0.5780	0.6045	0.6032	0.6099
	AUPRC*	RF	0.8048	0.8125	0.8260	0.8120	0.8144
		ANN	0.7504	0.7595	0.7530	0.7341	0.7450
		SVM	0.7537	0.7482	0.7261	0.7216	0.7331
		LR	0.7374	0.7435	0.7443	0.7437	0.7427
		KNN	0.7279	0.7342	0.7126	0.7076	0.7256
		DT	0.6849	0.6999	0.7196	0.7051	0.7199
Embedded (RFI)	Testing accuracy* (%)	RF	79.22	78.72	78.57	78.43	77.64
		ANN	73.36	73.43	73.86	72.07	66.43
		SVM	73.29	73.36	73.02	70.37	69.50
		LR	69.21	69.64	69.79	67.71	63.93
		KNN	75.64	72.64	73.93	72.00	72.50
		DT	69.43	68.50	68.21	70.07	70.93
	F1 score*	RF	0.7393	0.7358	0.7345	0.7307	0.7223
		ANN	0.6337	0.6328	0.6466	0.6156	0.4659
		SVM	0.6575	0.6564	0.6546	0.6186	0.5850
		LR	0.5559	0.5686	0.5739	0.5635	0.4626
		KNN	0.7178	0.6810	0.6987	0.6773	0.6828
		DT	0.6106	0.6084	0.6173	0.6463	0.6538
	AUPRC*	RF	0.8445	0.8498	0.8505	0.8374	0.8301
		ANN	0.7198	0.7084	0.7108	0.7279	0.6732
		SVM	0.7731	0.7678	0.7540	0.7624	0.6985
		LR	0.6922	0.6880	0.7122	0.6947	0.6050
		KNN	0.7735	0.7829	0.7865	0.7858	0.7764
		DT	0.6671	0.6901	0.6771	0.7041	0.6925

Bolded text indicated the best results achieved. Best model was selected based on AUPRC

*Average of testing accuracy, AUPRC, and F1 score from 10 times runs of 5-CV