Table 5.

Performance of reduced models with filter feature selection (PCC)

Performance		Number of features
		30	20	15	14	13	12
Testing accuracy* (%)	RF	72.07	70.64	70.79	69.29	67.36	66.36
	ANN	71.00	71.07	71.43	71.43	70.07	69.54
	SVM	69.64	71.00	69.29	69.14	66.79	66.64
	LR	70.64	70.21	71.00	70.50	70.07	68.93
	KNN	61.43	61.00	64.21	62.43	61.36	61.00
	DT	63.07	63.57	65.57	62.79	59.14	60.71
AUPRC*	RF	0.7383	0.7375	0.7450	0.6965	0.6784	0.6366
	ANN	0.7059	0.7215	0.6976	0.7190	0.6938	0.6516
	SVM	0.6823	0.6992	0.6465	0.6378	0.6177	0.6055
	LR	0.6906	0.6937	0.6907	0.6783	0.6791	0.6629
	KNN	0.6222	0.5962	0.5902	0.5653	0.5698	0.5424
	DT	0.6025	0.5949	0.6295	0.6157	0.5405	0.5806
F1 score*	RF	0.6609	0.6384	0.6450	0.6222	0.6015	0.5847
	ANN	0.6214	0.6278	0.6329	0.6275	0.6036	0.6099
	SVM	0.5867	0.6117	0.5927	0.5889	0.5557	0.5552
	LR	0.6078	0.6027	0.6125	0.5972	0.5890	0.5702
	KNN	0.4344	0.4355	0.5043	0.5040	0.4856	0.4851
	DT	0.5464	0.5699	0.5716	0.5524	0.4854	0.5072

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

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