In the machine learning of fuzzy rules for the diagnostic expert systems, the best selection of rule conditions is one of the most important steps. A new fuzzy nearness function was proposed, and the overlap degree of the rule conditions was evaluated to get an evaluated matrix with this function. An algorithm of the best selection for rule conditions or test points based on the evaluated matrix was designed. The simulation shows that the number of rule conditions, which has to be learned, is decreased largely, consequently, the workload of learning is reduced too. Further more, the method of the best selection of test points can do well to the design of automatic test.
Qi Yi, Shen Shituan, Li Yihua.Study for the best selection of rule conditions in automated extraction of fuzzy diagnostic rules[J] JOURNAL OF BEIJING UNIVERSITY OF AERONAUTICS AND A, 2004,V30(06): 506-511