Table 3.

Performance metrics of machine learning models without the Synthetic Minority Oversampling Technique.

Algorithms	Accuracy	Area under the receiver operating characteristic curve	Precision	Recall	Brier loss
Logistic regression	97.8	96.4	98.5a	93.4	0.016
Random forest	98.5b	97.7	98.1	96.1	0.012c
Extra trees	97.9	96.8	98.1	94.2	0.016
Gradient boosting trees	98.4	97.8d	97.5	96.5e	0.012c
Artificial neural network	97.1	95.3	96.6	91.8	0.024

^aThe highest precision.

^bThe highest accuracy.

^cThe least Brier loss.

^dThe highest area under the receiver operating characteristic curve.

^eThe highest recall.