Improving Retrieval Performance of Case Based Reasoning Systems by Fuzzy Clustering

Case-based reasoning (CBR), which is a classical reasoning methodology, has been put to use. Its application has allowed significant progress in resolving problems related to the diagnosis, therapy, and prediction of diseases. However, this methodology has shown some complicated problems that must be resolved, including determining a representation form for the case (complexity, uncertainty, and vagueness of medical information), preventing the case base from the infinite growth of generated medical information and selecting the best retrieval technique. These limitations have pushed researchers to think about other ways of solving problems, and we are recently witnessing the integration of CBR with other techniques such as data mining. In this article, we develop a new approach integrating clustering (Fuzzy C-Means (FCM) and K-Means) in the CBR cycle. Clustering is one of the crucial challenges and has been successfully used in many areas to develop innate structures and hidden patterns for data grouping [1]. The objective of the proposed approach is to solve the limitations of CBR and improve it, particularly in the search for similar cases (retrieval step). The approach is tested with the publicly available immunotherapy dataset. The results of the experimentations show that the integration of the FCM algorithm in the retrieval step reduces the search space (the large volume of information), resolves the problem of the vagueness of medical information, speeds up the calculation and response time, and increases the search efficiency, which further improves the performance of the retrieval step and, consequently, the CBR system.