Ensemble Learning Approach with Explainable AI for Improved Heart Disease Prediction

Ayomide Adekoya¹Faisal Saeed¹Wad Ghaban²Sultan Noman Qasem^3*

• ¹Birmingham City University, Birmingham, United Kingdom
• ²University of Tabuk, Tabuk, Saudi Arabia
• ³Imam Muhammad Ibn Saud Islamic University, Riyadh, Saudi Arabia

Heart disease remains a leading cause of global morbidity and mortality, motivating the development of predictive models that are both accurate and clinically interpretable. We introduce the Interpretable Ensemble Learning Framework (IELF), which integrates Explainable Boosting Machines (EBM) with XGBoost, SHAP-based explanations, and LIME for enhanced local interpretability. IELF was evaluated on two benchmark datasets: Cleveland (n=303) and Framingham (n=4,240). Across 5-fold cross-validation and held-out test sets, IELF achieved robust discrimination (AUC 0.899, accuracy 88.5% on Cleveland; AUC 0.696, accuracy 82.6% on Framingham) while maintaining balanced precision–recall trade-offs. Compared with EBM, IELF delivered significant improvements in recall, F1, and AUC on the Framingham dataset (p < 0.05), with differences versus XGBoost being less consistent. Beyond predictive performance, IELF provided transparent feature rankings aligned with established cardiovascular risk factors (e.g., age, blood pressure, cholesterol, diabetes, smoking), stable explanations across folds, and subgroup fairness analyses. To our knowledge, IELF is the first framework to combine EBM and XGBoost with SHAP and LIME, providing quantitative validation of explanation stability (Kendall’s τ, Overlap@10) under strict nested cross-validation with calibration and subgroup analyses. IELF further incorporates local feature explainability through LIME, offering complementary patient-level insights beyond SHAP. While headline accuracies were lower than recent state-of-the-art models reporting >97%, IELF was developed under stricter methodological controls (nested CV, calibration, subgroup validation), ensuring reproducibility, interpretability, and clinical reliability. These findings position IELF as a trustworthy benchmark for translational AI in cardiovascular risk prediction, complementing high accuracy but less transparent models.