Evaluating the Effectiveness of Feature Selection and Explainable AI in Predicting Acute Myocardial Infarction Using Machine Learning Models

Acute myocardial Infarctions (MIs), commonly known as heart attacks, are critical medical emergencies that often lead to severe complications, including post-myocardial infarction angina (PMIA). Early and accurate prediction of PMIA is crucial for timely intervention and improved patient outcomes. This study focuses on predicting PMIA using machine learning classifiers on the UCI myocardial infarction complication datasets, which includes 1700 patient samples with 111 clinical features. The dataset is balanced using Borderline Synthetic Minority Oversampling (Borderline-SMOTE) and standardized. Twelve classifiers are employed and the Shapley Additive exPlanations (SHAP) are used to interpret the best-performing models, revealing the decision-making processes. The SVM classifier achieved the highest cross-validated accuracy of 98.55%, demonstrating superior predictive capability. SHAP values provided insights into the importance of each feature, enhancing model interpretability and transparency. This approach highlights the significance of feature selection and explain ability in building trustworthy AI models for healthcare. The findings contribute to improved diagnosis and management of PMIA, underscoring the ethical adoption of AI in clinical settings.