XAI-Augmented Decision Support Systems for Complex, Multi-criteria Problems in Uncertain Healthcare Environments

XAI-Augmented Decision Support Systems for Complex, Multi-criteria Problems in Uncertain Healthcare Environments

연구 분야: Verification

학회: International Conference on Science, Engineering Management and Information Technology

Under uncertainty, healthcare decisions often involve conflicting criteria (e.g., diagnosis accuracy, patient risk, treatment cost). While data-driven models offer strong predictive performance, their opacity limits clinician trust and adoption. We present a cloud-based, Java-implemented decision support system (DSS) that combines multi-criteria decision analysis (MCDA) with explainable AI (XAI) techniques—SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations). The system enhances transparency by generating local explanations for black-box model outputs. Its modular framework spans data preprocessing, machine learning, explanation generation, and multi-criteria evaluation. A real-world case study, such as disease risk prediction, demonstrates how SHAP/LIME visualizations increase clinical interpretability. The MCDA layer aggregates conflicting objectives into an interpretable decision score, supporting more informed choices. We evaluate the system using conventional metrics (accuracy, precision, recall, F1) and explanation integrity measures, showing that XAI improves user understanding without sacrificing performance. Future directions include real-time inference, richer uncertainty modeling, and guidance on integrating XAI into healthcare DSS workflows.