Cardiovascular disease (CVD) remains the leading cause of mortality and morbidity worldwide, highlighting the necessity for early and accurate risk stratification methods that can improve patient outcomes. This chapter proposes a comparative evaluation of nine supervised machine learning algorithms for the classification of CVD severity. Using the online UCI Heart Disease dataset, each algorithm is evaluated across multiple configurations: a baseline model, three models employing different oversampling techniques, one model with optimized hyperparameters, and finally, the optimized models combined with oversampling. The presented research findings demonstrate that ensemble-based approaches, particularly CatBoost and XGBoost consistently achieve higher predictive performance than traditional classifiers, especially when combined with oversampling techniques. In addition to performance optimization, this chapter emphasizes explainability to enhance the interpretability of model predictions. SHapley Additive exPlanations are utilized to facilitate clinical integration by improving the transparency of model decision-making. Overall, this study highlights the importance of technical performance and model interpretability in the development of clinically relevant support systems for cardiovascular diagnostics.