Learning from Successes and Failures: An Exploration of a Case-Based Reasoning Technique
Résumé
Usually, existing works on adaptation in case-based reasoning assume that the case base holds only successful cases, i.e., cases having solutions believed to be appropriate for the corresponding problems. However, in practice, the case base could hold failed cases, resulting from an earlier adaptation process but discarded by the revision process. Not considering failed cases would be missing an interesting opportunity to learn more knowledge for improving the adaptation process. This paper proposes a novel approach to the adaptation process in the case-based reasoning paradigm, based on an improved barycentric approach by considering the failed cases. The experiment performed on real data demonstrates the benefit of the method considering the failed cases in the adaptation process compared to the classical ones that ignore them, thus, improving the performance of the case-based reasoning system.