EXPLAINABLE AI-BASED HEART DISEASE PREDICTION USING SPARSEGPT-INSPIRED XGBOOST AND LIME FOR RISK FACTOR PRIORITIZATION IN YOUNG ADULTS.

Abstract

Cardiac arrest and cardiovascular disorders among young adults are rapidly increasing due to complex interdependencies among clinical, lifestyle, and genetic factors. Conventional diagnostic models often struggle to capture these nonlinear relationships while maintaining transparency and computational efficiency essential for real-world deployment. This study proposes a SparseGPT-inspired XGBoost framework integrated with Explainable Artificial Intelligence (XAI) using Local Interpretable Model-Agnostic Explanations (LIME) for accurate, interpretable, and resource-efficient cardiac health prediction. The model incorporates L₂ regularization-based structured pruning, mimicking SparseGPT’s sparsity principles to significantly reduce parameter redundancy and computational load while preserving high predictive accuracy. A comprehensive dataset comprising clinical records, lifestyle indicators, physiological parameters, and genetic predispositions of young adults was curated, normalized, and balanced to ensure robustness. The optimized model achieved an accuracy of 98.12%, F1-score of 0.98, and an R² value of 0.97, outperforming conventional ensemble and deep learning baselines. Through LIME-based feature interpretability, key determinants such as serum cholesterol, resting blood pressure, chest pain type, age, and exercise-induced angina were identified and ranked, enabling clinically relevant risk factor prioritization. Furthermore, the model’s lightweight design facilitates seamless integration into edge and wedge computing ecosystems, allowing real-time cardiac risk assessment through wearable or IoT-enabled health monitoring devices. The proposed hybrid framework thus bridges the gap between predictive intelligence and clinical transparency, providing a trustworthy, explainable, and computationally efficient solution for next-generation preventive cardiology and intelligent healthcare systems.