TABLE 1.

Optimal classification performance (AUC-ROC, BA, and MCC values) of each feature selection method for predicting KS using a cross-validation of training set results.

Feature selection method	Machine learning algorithm	Rebalancing method	Cutoff	AUC-ROC	BA	MCC
AUC-ROC	NB	ROSE	0.52	0.72 ± 0.03	0.69 ± 0.02	0.27 ± 0.02
AUC-ROC	NNET	Original	0.52	0.70 ± 0.03	0.68 ± 0.02	0.28 ± 0.03
AUC-ROC	RF	Upsampling	0.52	0.76 ± 0.02	0.72 ± 0.02	0.32 ± 0.03
AUC-ROC	SVM	Upsampling	0.52	0.74 ± 0.02	0.70 ± 0.02	0.29 ± 0.03
AUC-ROC	XGBoost	Original	0.52	0.76 ± 0.02	0.71 ± 0.02	0.31 ± 0.03
Fisher’s exact test	NB	Original	0.01	0.78 ± 0.01	0.73 ± 0.01	0.33 ± 0.02
Fisher’s exact test	NNET	ROSE	0.05	0.74 ± 0.02	0.70 ± 0.02	0.28 ± 0.04
Fisher’s exact test	RF	Upsampling	0.05	0.81 ± 0.01	0.75 ± 0.01	0.35 ± 0.02
Fisher’s exact test	SVM	Upsampling	0.03	0.78 ± 0.01	0.73 ± 0.01	0.32 ± 0.02
Fisher’s exact test	XGBoost	Upsampling	0.05	0.79 ± 0.01	0.73 ± 0.01	0.32 ± 0.03
RF	NB	Original	40	0.74 ± 0.02	0.69 ± 0.02	0.28 ± 0.03
RF	NNET	Upsampling	50	0.69 ± 0.02	0.66 ± 0.01	0.23 ± 0.03
RF	RF	Original	50	0.77 ± 0.02	0.71 ± 0.02	0.32 ± 0.04
RF	SVM	Upsampling	50	0.73 ± 0.02	0.69 ± 0.02	0.28 ± 0.04
RF	XGBoost	Original	50	0.75 ± 0.02	0.70 ± 0.02	0.30 ± 0.04
XGBoost	NB	Original	50	0.76 ± 0.02	0.71 ± 0.02	0.30 ± 0.03
XGBoost	NNET	Original	40	0.72 ± 0.02	0.69 ± 0.02	0.29 ± 0.03
XGBoost	RF	Original	50	0.79 ± 0.02	0.74 ± 0.02	0.35 ± 0.04
XGBoost	SVM	Upsampling	50	0.76 ± 0.02	0.71 ± 0.02	0.30 ± 0.02
XGBoost	XGBoost	Upsampling	50	0.77 ± 0.02	0.71 ± 0.02	0.30 ± 0.02