Learning and Applying Range Adaptation Rules in Case-Based Reasoning Systems

The retrieval-only Case-Based Reasoning (CBR) systems do not provide acceptable accuracy in critical domains such as medical. Besides, the case adaptation process in CBR is often a challenging issue as it has been traditionally carried out manually by domain experts. In this paper, a new case-based approach using transformational adaptation rules called "range adaptation rules" is proposed to improve the accuracy of a retrieval-only CBR system. The rangeadaptation rules are automatically generated from the case-base. In this approach, after solving each new problem, the case-base is expanded and the range adaptation rules are updated automatically. To evaluate the proposed approach, a prototype is implemented and experimented in agriculture domain to classify the IRIS plant types. The experimental results show that the proposed approach increases the classification accuracy comparing with the retrieval-only CBR system.