Enhancing Case Retrieval in Case-Based Reasoning Through Improved Solution Space Diversity and Coverage

Case-Based Reasoning (CBR) is a well-established methodology used in Systems Engineering as a decision support tool. However, in large case bases containing numerous similar cases, the retrieval process often yields solutions with low diversity, limiting the usefulness of the system in complex decisionmaking scenarios. This work introduces a novel approach to case base maintenance that enhances diversity while preserving retrieval effectiveness. The proposed method separates the description and solution spaces and applies a modified Condensed Nearest Neighbor (CNN) algorithm to generalize and reindex similar cases. Rather than deleting redundant cases, the approach integrates them into parent-child structures, maintaining a wide range of solutions while reducing redundancy in descriptions. A case study in predictive maintenance system design demonstrates the method's effectiveness. Results show that the case base size can be reduced by 82.14 %, while improving the diversity of retrieved solutions by 132.96 % and maintaining over 95 % of the original coverage. This approach supports more robust and diverse retrieval outcomes, ultimately enhancing decision support capabilities. The method offers a scalable and efficient solution to the challenge of diversity in CBR, making it a valuable contribution to Systems Engineering and other domains where knowledge reuse is critical.