Prediction of incident myocardial infarction using machine learning applied to harmonized electronic health record data

Table 3 Comparison of machine learning approaches

	F1 Score	AUC	Training time
Naïve Bayes	0.060	0.73	1 min
Logistic regression with L2 regularization	0.084	0.829	1 min
Logstic regression with no regularization	0.06	0.79	1 min
RF	0.084	0.765	3 min
Shallow NN	0.101	0.83	1 min
Deep NN	0.092	0.835	2 min
GBM	0.077	0.83	9 min

Comparison of various models using Random Undersampling technique and all features. F1 and AUC calculated from model applied to held-out testing set (20%); training time is for training of training set (80%)

ISSN: 1472-6947