Learning a coverage set of maximally general fuzzy rules by rough sets

Expert systems have been widely used in domains where mathematical models c annot be easily built, human experts are not available or the cost of query ing an expert is high. Machine learning or data mining can extract desirabl e knowledge or interesting patterns from existing databases and ease the de velopment bottleneck in building expert systems. In the past we proposed a method [Hong, T.P., Wang, T.T., Wang, S.L. (2000). Knowledge acquisition fr om quantitative data using the rough-set theory. Intelligent Data Analysis (in press).], which combined the rough set theory and the fuzzy set theory to produce all possible fuzzy rules from quantitative data. In this paper, we propose a new algorithm to deal with the problem of producing a set of m aximally general fuzzy rules for coverage of training examples from quantit ative data. A rule is maximally general if no other rule exists that is bot h more general and with larger confidence than it. The proposed method firs t transforms each quantitative value into a fuzzy set of linguistic terms u sing membership functions and then calculates the fuzzy lower approximation s and the fuzzy upper approximations. The maximally general fuzzy rules are then generated based on these fuzzy approximations by an iterative inducti on process. The rules derived. can then be used to build a prototype knowle dge base in a fuzzy expert system. (C) 2000 Elsevier Science Ltd. All right s reserved.