Improving Real-Time Efficiency of Case Retrieving Process for Case-Based Reasoning

Conventional case-based reasoning (CBR) does not perform efficiently for high-volume datasets because of case retrieval time. To overcome this problem, previous research suggested clustering a case base into several small groups and retrieving neighbors within a corresponding group to a target case. However, this approach generally produces less accurate predictive performance than the conventional CBR. This paper proposes a new case-based reasoning method called the clustering–merging CBR (CM-CBR). The CM-CBR method dynamically indexes a search pool to retrieve neighbors considering the distance between a target case and the centroid of a corre-sponding cluster. This method is applied to three real-life medical datasets. Results show that the proposed CM-CBR method produces similar or better predictive performance than the conventional CBR and clustering-CBR methods in numerous cases with significantly less computational cost.