Relevance. To date, the spread of myocardial infarction (MI) accounts for 38.9 deaths per 100 thousand people. To reduce this indicator, it is important to reduce the number of post-hospital complications in patients after discharge. For such purposes, forecast scales are actively used today: EDACS-ADP, HEART, GRACE, etc. However, recently many scientists have been talking about the possibility of using artificial intelligence technology in order to improve the quality of the forecast. The goal is to develop and test new methodological approaches for predicting deaths for 6 months after discharge from the hospital in patients who have undergone MI using machine learning technology. Materials and methods: the work was carried out on the data of the international study "Acute myocardial infarction in the Russian Federation: current practice and obstacles to effective treatment at different levels of the healthcare system", n=1,128: men n=845 (74.9%); women n=248 (21.9%). Both cohorts were divided into 2 groups: First group n= 980 (86.9%) is patients who survived for a 6–month follow-up period after discharge from the hospital. Second group 2 n=65 (5.8%) is patients who died during the 6-month follow-up period after discharge from the hospital, including deaths in the hospital. To build the prediction model, we used six classification algorithms. Predictors included in the analysis are n=193. To evaluate the effectiveness of forecast models, we used ROC analysis indicators. Results. The best algorithm turned out to be LGBMClassifier AUC-0.84, the worst results were shown by logistic regression AUC-0.79. The most significant factors were the observation of a therapist and a cardiologist during the year, the employment of patients and glomerular filtration rate indicators. Risk factors such as medical examination, a history of chronic heart failure and comorbidity also played an important role. Discussion. As a result of the discussion, it was possible to compare the most important risk factors affecting post-hospital mortality and compare them with those obtained as a result of the experiment. We also made sure that the average AUC value obtained in similar studies – 0.82 corresponds to the results we obtained. Conclusions. As a result of the study, we were able to construct a classification model with a fairly good quality of determination. Further studies of this topic will help to better understand the nature of pathological changes development and to prevent adverse outcomes in advance.