Explainable Predictive Analytics for Healthcare Decision Support

Abstract

Predictive analytics has become a cornerstone of data-driven healthcare, enabling early diagnosis, disease risk assessment, and personalized treatment recommendations. However, the “black-box” nature of many machine learning models limits clinical trust and interpretability, which are essential for real-world adoption. This study presents a comprehensive framework for Explainable Predictive Analytics (XPA) that integrates interpretable machine learning techniques with explainability tools such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) to support transparent healthcare decision-making. Using benchmark datasets on heart disease and diabetes, the proposed framework improves both predictive accuracy and interpretability, achieving up to 93% accuracy while maintaining high feature attribution consistency across models. The results demonstrate that integrating explainability mechanisms not only enhances model transparency but also empowers clinicians with actionable insights for diagnostic and treatment support.