Many fundamental features of diagnostic reasoning have been disclosed by rescent researches in decision science, cognitive science, and artificial intelligence. Production rules (if-then rules) is one of the most popular tools for constructing computer-based diagnostic systems or consultation systems. However, a major difficulty in the design of rule-based expert system is the process of acquiring the requisite knowledge in the form of if-then rules. In most of rule-based expert systems, crispy or fuzzy if-then rules generally derived from human experts using linguistic rules are invariably rather crude and, although qualitatively correct, need to be refined to achieve better performance. In this paper, we present an innovative approach to the rule extraction directly from experimental numerical data. This paper presents a novel class of hyperrectangular composite neural networks (HRCNN's) which are trained in such a way that they provide an appealing solution to the problem of knowledge acquisition. The parameters of HRCNN's, after sufficient training, are then utilized to generate both crispy and fuzzy if-then rules. The ultimate goal of this neuro-fuzzy approach is to free diagnosticians from tedious diagnostic loads. We apply the proved proposed HRCNN's to the example of a hypothesis regarding the pathophysiology of diabetes. From the experimental results, it is summarized that the performance was rather acceptable.
|頁（從 - 到）||138-144|
|期刊||Biomedical Engineering - Applications, Basis and Communications|
|出版狀態||已出版 - 1 1月 1996|