Clustering cases for case-based reasoning systems

Case-based reasoning is a new approach to develop knowledge-based systems. Since cases are stored in the secondary storage, their retrievals and insertions directly determine system's response time. Therefore, we need a case clustering approach to organize cases in disk pages for fast retrievals and insertions. This paper proposes a new case clustering approach. The advantage of the new approach is that it relaxes a restrictive assumption made by a previous approach. Also, the new approach is more flexible and uses either an optimal or a heuristic clustering procedure. Preliminary results from our performance evaluation indicate that the heuristic procedure is not only efficient but also produces optimal solutions to problems for which optimal solutions are found by the optimal procedure.