Improving case retrieval performance through the use of clustering techniques

Abstract

The performance of Case-Based Reasoning (CBR) systems is highly depend on the performance of the retrieval phase. Usually, if the case memory has a large number of cases the system turn to be very slow. Several mechanisms have been proposed in order to prevent a full search of the case memory during the retrieval phase. In this work we propose a clustering technique applied to the memory of cases. But this strategy is applied to an intermediate level of information that defines paths to the cases. Algorithms to the retrieval and retention phase are also presented.