Table 2.

Performance of machine learning algorithms for predicting 30-day mortality

Model name	F1	ROC AUC	Recall	Brier score
XGBoost	0.65	0.78	0.84	0.25
SVM	0.64	0.75	0.63	0.19
Decision tree	0.59	0.72	0.61	0.23
Gradient boosting	0.58	0.72	0.56	0.22
Logistic regression	0.58	0.71	0.51	0.21
Random forest	0.55	0.69	0.46	0.20
KNN	0.27	0.55	0.20	0.29

F1 score is the harmonic mean of the precision (positive predictive value) and the recall (sensitivity), where higher scores are better. ROC AUC, receiver operator characteristic area under the curve, which is a marker of discrimination between classes and higher scores are preferable. The Brier score is a marker of calibration and discrimination, where lower scores are better.

KNN, k-nearest neighbours; SVM, support vector machines.