This work focuses on solving the problem of predicting the fatal outcome after a myocardial infarction based on a large number of factors obtained during the survey and examination of patients. The prognosis of patients undergoing a heart attack varies and depends on many factors. Identification of the most important of them that significantly affect the lethal outcome is one of the most important tasks of cardiology. To solve this classification problem, various machine learning algorithms were consistently applied. Methods such as gradient boosting, random forest, and others were applied to the existing database of patients who had suffered a myocardial infarction. The most informative features were also selected so that the doctor could get a prognosis using only a few features.