This paper proposes a genetic-algorithm-based approach to theconstruction of fuzzy classification systems with rectangular fuzzyrules. In the proposed approach, compact fuzzy classification systemsare automatically constructed from numerical data by selecting a smallnumber of significant fuzzy rules using genetic algorithms. Sincesignificant fuzzy rules are selected and unnecessary fuzzy rules areremoved, the proposed approach can be viewed as a knowledge acquisitiontool for classification problems. In this paper, first we describe ageneration method of rectangular fuzzy rules from numerical data forclassification problems. Next, we formulate a rule selection problem forconstructing a compact fuzzy classification system as a combinatorialoptimization problem. Then we show how genetic algorithms are applied tothe rule selection problem
展开▼
