The study developed and validated machine learning (ML) models to predict mortality after heart valve surgery. The researchers analyzed data from 935 patients and compared the performance of the ML models with the conventional EuroSCORE II scoring system. Machine learning models, especially Extra Trees and logistic regression, showed better accuracy in predicting in-hospital and 30-day postoperative mortality compared to EuroSCORE II. For 365-day mortality, the predictive performance of the models was comparable. Key predictors included age and biomarkers of heart, kidney, and liver function. Based on the best models, nomograms were created that serve as practical tools for estimating individual risk after valve surgery. The study suggests that ML models can improve risk stratification and aid decision-making in clinical practice.