A Comprehensive Review of Case Representation and Similarity Measures in Case-Based Reasoning Systems

Abstract

Case-Based Reasoning (CBR) is a human-inspired problem-solving approach where new problems are solved by recalling and adapting solutions from similar past cases. The performance of a CBR system critically depends on how cases are represented and how similarity between cases is computed. These two factors determine the accuracy, efficiency and applicability of CBR systems across diverse domains. This paper presents a comprehensive and comparative review of various case representation techniques and similarity measures. The review evaluates these methods based on important measures such as interpretability, scalability, adaptability, computational complexity and retrieval effectiveness. It further explores their suitability across domains including healthcare, finance, engineering and disaster management. The analysis reveals that no single technique is universally optimal; rather, the alignment between representation format and similarity computation, often through hybridization or domain-specific adaptation, is critical to achieving optimal system performance. Through rich literature insights and practical illustrations, the paper identifies emerging trends such as machine learning-driven similarity adaptation, ontology automation and real-time retrieval, offering a roadmap for the next generation of intelligent and context-aware CBR systems.