Explainable AI-Enhanced Machine Learning Models for Coronary Heart Disease Detection

Coronary Heart Disease (CHD) continues to be the leading cause of death worldwide, underscoring the necessity for accurate early detection and risk assessment tools. This study uses a dataset of 303 patients with 14 clinically significant features to examine the use of machine learning (ML) techniques for CHD prediction. Comprehensive preprocessing was employed, including BorderlineSMOTE for class imbalance correction and grid search for hyperparameter optimization, ensuring robust model performance. Multiple classifiers, including Random Forest, Logistic Regression, and XGBoost, were evaluated. While Random Forest and Logistic Regression achieved accuracies of 84% and 79%, respectively, the stacked ensemble model demonstrated superior performance, attaining an accuracy of 86% and an AUC-ROC of 0.90. Explainable AI (XAI) techniques, such as SHAP and LIME, were employed to enhance interpretability. These methods identified critical predictive features, including slope, thall, and oldpeak, which are known indicators of ischemic burden and myocardial stress key factors in CHD assessment. By offering transparent, interpretable insights, these techniques facilitate clinician trust and support AI integration into routine diagnostics. Despite promising results, the study's findings are based on a relatively small dataset of 303 patients, warranting further validation on larger, more diverse datasets to improve model generalizability. To improve clinical applicability and predictive accuracy, future studies should investigate integration with real-time monitoring systems and hybrid deep learning (DL) techniques. With the potential to enhance patient outcomes and early intervention, this study adds to the expanding field of AI-driven cardiovascular diagnostics.