Abstract: Heart Cardiovascular disease continues to be one of the leading causes of death globally, especially in under-served and rural communities where access to sophisticated diagnostic tools is limited. This study explores the use of machine learning (ML) and explainable artificial intelligence (XAI) to provide accessible, reliable, and interpretable early detection of heart disease. Leveraging a synthetically generated dataset modelled on common patient profiles based on features like age, cholesterol, chest pain type, ECG readings, and maximum heart rate we developed and evaluated four ML models: Logistic Regression, Random Forest, Gradient Boosting, and XG Boost.

Our work proposes a two-tier diagnostic framework: a lightweight, mobile-friendly model for community-level screening, and a more advanced model for clinical environments. We employed SHAP (SHapley Additive exPlanations) to ensure model interpretability and transparency, critical for clinical adoption. The results are promising, with the mobile-tier model achieving 81% accuracy and the clinical-tier model reaching 89%. These findings underscore the potential of interpretable AI to democratize cardiac care, particularly in areas lacking medical infrastructure. Future directions include integrating wearable devices and telemedicine to support real-time monitoring and broader health equity.

| DOI: 10.17148/IARJSET.2025.12627