Dynamic case base maintenance for a case-based reasoning system

Abstract

The success of a case-based reasoning system depends critically on the relevance of the case base. Much current CBR research focuses on how to compact and refine the contents of a case base at two stages, acquisition or learning, along the problem solving process. Although the two stages are closely related, there is few research on using strategies at both stages at the same time. This paper presents a model that allows to update itself dynamically taking information from the learning process. Different policies has been applied to test the model. Several experiments show its effectiveness in different domains from the UCI repository. © Springer-Verlag Berlin Heidelberg 2004.