Table 5.

Statistics on AUCs grouped by different factors.

Grouping factors	Groups	N	Mean	St. Dev.	Min	Pctl(25)	Pctl(75)	Max	Range
Feature selection strategy	Original strategy	75	0.6547	0.0428	0.5125	0.6372	0.6819	0.7292	0.2167
	LASSO-based strategy	75	0.6582	0.0441	0.4948	0.6449	0.6821	0.7239	0.2291
	Forward selection strategy	75	0.6426	0.0626	0.5085	0.6307	0.6846	0.7355	0.2270
	Backward selection strategy	75	0.6580	0.0427	0.5152	0.6435	0.6788	0.7331	0.2179
	Importance-based strategy	75	0.6550	0.0436	0.5125	0.6350	0.6823	0.7307	0.2182
Machine learning model	RF	75	0.6889	0.0314	0.6100	0.6713	0.7129	0.7355	0.1255
	LR	75	0.6666	0.0222	0.6297	0.6514	0.6792	0.7161	0.0864
	XGBoost-tree	75	0.6574	0.0256	0.6110	0.6353	0.6748	0.7171	0.1061
	NNET	75	0.6552	0.0259	0.6163	0.6362	0.6702	0.7239	0.1076
	SVM-linear	75	0.6005	0.0668	0.4948	0.5360	0.6558	0.7184	0.2236
Sampling method	Oversampling	125	0.6604	0.0403	0.5085	0.6385	0.6908	0.7355	0.2270
	Undersampling	125	0.6548	0.0354	0.5085	0.6398	0.6756	0.7239	0.2154
	Original sampling	125	0.6460	0.0626	0.4948	0.6365	0.6911	0.7340	0.2392

N: number of cases. Mean: mean value corresponding to AUC of each model. St. Dev.: standard deviation corresponding to AUC of each model. Pctl(25): AUC corresponds to the first quartile of the variance numerical distribution of each model. Pctl(75): AUC corresponds to the third quartile of the variance numerical distribution of each model. LR: logistic regression model. NNET: neural networks. RF: random forest. SVM-linear: support vector machine-linear. XGBoost-tree: extreme gradient boosting-tree.