MedRule: An Interpretable Rule Induction Framework for Reliable Clinical Decisions

The use of artificial intelligence in healthcare decision support is expanding, but many systems still operate as opaque black boxes. In clinical practice, limited transparency can reduce confidence and restrict the usefulness of such tools. This paper introduces MedRule, a framework for medical decision support that prioritizes interpretability and reliability. The approach leverages class-association rule mining to generate decision paths that are clear and clinically meaningful. The framework represents knowledge through concise if–then rules, enabling clinicians to examine and validate how conclusions are reached. A dedicated induction strategy ensures that the extracted rules remain both coherent and manageable, supporting daily use without obscuring reasoning. Experimental evaluation on healthcare datasets indicates that MedRule achieves competitive predictive performance while generating compact and clinically meaningful rule sets, thus enhancing interpretability and supporting trustworthy AI-assisted clinical decision-making.