Adaptive case-based reasoning using support vector regression

One important step in case-based reasoning systems is the adaptation phase. This paper presents a case-based reasoning system which automatically adapts past solutions to propose a solution for new problems. The proposed method for case adaptation is based on support vector regression. At first, case base is partitioned using SOM technique. Then, a support vector regression is constructed for each cluster using local information. For solving a new problem, its local information is computed with respect to the most similar cluster and the corresponding support vector regression propose a solution. Experiment shows this approach greatly improves the accuracy of a retrieve-only CBR system with minimizing each didactic model.