Predicting Mortality in Patients with Chronic Heart Failure

Abstract

Chronic heart failure affects many people worldwide. A random forest model was created in this study to help predict if the patients die or survive. The model was based on the data from MUSIC database and on data extracted from highresolution ECG records in the database. In total, 40 features were used for the training. The model achieved 0.50 F1 score, 0.73 AUROC and 0.60 AUPRC on test set. The most significant features were ”Pro BNP (ng/L)” and ”Urea (mg/dL)”.