Reviewing The Combination of Case-Based Reasoning and Machine Learning for Improving a Decision Support System

Abstract

This study investigates the integration of case-based reasoning (CBR) with machine learning (ML) to enhance decision support systems. Due to the inadequate synthesis of empirical data in this domain by previous research, we undertook a systematic review of 46 indexed journal articles published between 2019 and 2024. The review adhered to PRISMA principles to ensure a transparent and rigorous selection and analysis procedure. We analyzed integration architectures and documented performance results, application areas, and persistent implementation challenges. The research indicates that hybrid CBR-ML systems typically surpass single-method systems in accuracy, precision, and flexibility, with an average improvement of approximately 7% over CBR-only systems. Sequential and ensemble approaches typically demonstrate effectiveness, though weighted hybrid designs often achieve superior precision and recall, particularly in complex problem domains such as healthcare and finance where accuracy is critical. Researchers based in Asia authored the majority of the reviewed studies, with contributions from Europe and Africa following. These regions concentrated high-impact applications in healthcare, finance, manufacturing, and environmental management. Notwithstanding these developments, several enduring challenges persist, including substantial computational demands, vulnerability to variations in data quality, and the continuing scarcity of clearly articulated evaluation protocols, which hinder the effective implementation of high-impact applications across these regions. Overall, existing findings suggest that integrating reasoning-oriented approaches with learning-based methods can yield a balanced trade-off between predictive accuracy and interpretability.