Stacking Ensemble Models Predict Mortality from Acute Myocardial Infarction

In this study, a stacking ensemble model is proposed to predict in-hospital mortality for NSTEMI. The stacking ensemble model uses a two-tier structure to improve the performance of the algorithm. The first layer of the model uses decision trees (DT), support vector machines (SVM), random forests (RF), gradient boosting decision trees (GBDT) as the base model, and logistic regression (LR) is selected as the metamodel in the second layer. By comparing with four machine learning models, the results of experiments show that the AUC of the stacking model is (0. 958), higher than DT (0.934), SVM (0.942), RF (0.945), GBDT (0.948). In terms of AUC, Accuracy, Precision, Recall, and F1, the indicators improved by 1%, 1%, 0.7%, 0.4% and 0.9%, respectively, compared with the highest of the four single models.