TABLE 2.

Comparison of performance between risk prediction models.

Models	AUROC	MCC	Accuracy	Sensitivity	Specificity	PPV	NPV	F1
CatBoost	0.796	0.448	0.746	0.783	0.674	0.825	0.614	0.803
XGBoost	0.796	0.399	0.724	0.767	0.641	0.807	0.584	0.786
LightGBM	0.789	0.403	0.724	0.761	0.652	0.811	0.583	0.785
Random forest	0.742	0.413	0.728	0.761	0.663	0.815	0.587	0.787
Gradient boost	0.732	0.371	0.710	0.750	0.630	0.799	0.563	0.774
Linear SVM	0.721	0.373	0.699	0.706	0.685	0.814	0.543	0.756
MLP	0.728	0.379	0.710	0.739	0.652	0.806	0.561	0.771
CAD consortium clinical	0.727	0.313	0.676	0.706	0.620	0.784	0.518	0.743
CAD consortium basic	0.715	0.223	0.559	0.444	0.783	0.800	0.419	0.571
Diamond-Forrester score	0.687	0.271	0.706	0.933	0.261	0.712	0.667	0.808
K-nearest neighbor	0.704	0.313	0.676	0.706	0.620	0.784	0.518	0.743

AUROC, area under the receiver operating characteristics; CAD, coronary artery disease; GBM, gradient boosting machine; MCC, Matthews correlation coefficients; NPV, negative predictive value; PPV, positive predictive value; SVM, support vector machine; XG, extreme gradient. The bold values indicate the best performance of the 11 models.