Acute myocardial infarction is a major cause of hospitalization and mortality in China, where ST-elevation myocardial infarction (STEMI) is more severe and has a higher mortality rate. Accurate and interpretable prediction of in-hospital mortality is critical for STEMI patient clinical decision making. In this study, we used interpretable machine learning approaches to build in-hospital mortality prediction models for STEMI patients from Chinese Acute Myocardial Infarction (CAMI) registry data. We first performed cohort construction and feature engineering on CAMI data to generate an available dataset and identify potential predictors. Then several supervised learning methods with good interpretability, including generalized linear models, decision tree models, and Bayes models, were applied to build prediction models. The experimental results show that our models achieve higher prediction performance (AUC = 0.80~0.85) than the previous in-hospital mortality prediction STEMI models and are also easily interpretable for clinical decision support.