TABLE 2.

Performance Metrics of Machine Learning Models

Model	Dataset	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	TPP (n)	AUC
LR	Training*	7.2±0.7	99.6±0.1	50.2±7.4	95.4±0.2	73.6±16.5	0.842±0.011
	Testing	9.7	99.6	53.3	95.7	107	0.836
RF	Training	15.0±0.8	99.2±0.1	50.3±2.6	95.7±0.2	149.6±13.0	0.859±0.007
	Testing	15.1	99.1	46.1	96.0	193	0.848
ANN	Training	18.8±10.2	98.9±0.8	50.0±7.3	95.9±0.6	202.8±121.6	0.867±0.005
	Testing	3.6	99.9	53.8	95.5	39	0.845
SGD	Training	4.2±2.5	99.8±0.2	50.0±7.2	95.2±0.4	43.0±26.3	0.826±0.010
	Testing	2.2	100.0	72.2	95.4	18	0.819
NB	Training	8.5±1.3	98.9±0.3	28.4±3.4	95.4±0.2	151.8±33.5	0.824±0.011
	Testing	8.3	98.5	21.0	95.6	223	0.800
SVM	Training	0.1±0.1	100±0.0	16.7±21.1	95.1±0.3	1.6±0.8	0.755±0.010
	Testing	0.2	100	25	95.3	4	0.729
DT	Training	24.2±1.5	95.0±0.3	20.1±0.9	96.0±0.3	600.4±37.0	0.596±0.008
	Testing	24.9	94.6	18.5	96.3	793	0.598

Data are presented as mean±SD unless indicated otherwise.

^*Performance metrics are obtained from 5-fold cross-validation.

ANN indicates artificial neural network; AUC, area under the receiver operating characteristic curve; DT, decision tree; LR, logistic regression; NB, naive Bayes; NPV, negative predictive value; PPV, positive predictive value; RF, random forest; SGD, stochastic gradient descent; SVM, support vector machine; TPP (n), total positive prediction number.