Table 2.

Comparison of model performance in training, testing, and validation data sets

	Logistic Regression		Naïve Bayes		Random Forest		Support Vector
Training set (cross-validation)
AUC	0.714	0.037	0.716	0.039	0.706	0.037*	0.697	0.042*
Specificity	0.867	0.031	0.783	0.035*	0.886	0.032^	0.917	0.041^
Sensitivity	0.353	0.058	0.519	0.064^	0.336	0.053*	0.217	0.068*
Testing set
AUC	0.712	0.023	0.698	0.024	0.693	0.024	0.706	0.023
Specificity	0.870	0.019	0.781	0.023	0.889	0.016	0.920	0.015
Sensitivity	0.332	0.033	0.492	0.036	0.337	0.033	0.241	0.029
	X2	p val	X2	p val	X2	p val	X2	p val
Calibration	4.09	0.5	102.73	<.001	5.04	0.4	17.02	0.005
Validation set
AUC	0.719	0.024
Specificity	0.911	0.016
Sensitivity	0.259	0.036
	X2	p val
Calibration	2.38	0.8

^{^}, superior to logistic regression (p<0.05)

^*, inferior to logistic regression (p<0.05)

AUC, area under receiver operating characteristic curve