A Retrieval Strategy for Case-Based Reasoning Using Similarity and Association Knowledge

摘要

Retrieval is a key phase in case-based reasoning (CBR), since it lays the foundation for the overall effectiveness of CBR systems. Its aim is to retrieve useful cases that can be used to solve the target problem. To perform the retrieval process, CBR systems typically exploit similarity knowledge and is called similarity-based retrieval (SBR). However, SBR tends to rely strongly on similarity knowledge, ignoring other forms of knowledge that can be further leveraged to improve the retrieval performance. This paper argues and motivates that association analysis of stored cases can significantly strengthen SBR. We propose a novel retrieval strategy USIMSCAR that substantially outperforms SBR by leveraging association knowledge, encoded via a certain form of association rules, in conjunction with similarity knowledge. We also propose a novel approach for extracting association knowledge from a given case base using various association rule mining techniques. We evaluate the significance of USIMSCAR in three application domains—medical diagnosis, IT service management, and product recommendation.