This paper proposes a GA-based method to construct fuzzy classification system with high precision and small number of fuzzy rules and input variables.The method can be divided into three steps. In the first step, the antecedents of fuzzy classification system and input variables are coded into a binary string and treated as an individual in genetic algorithm. The fitness function is specified by correctly classified patterns, the number of fuzzy rules and input variables. The initial fuzzy classification system is constructed by binary-coded genetic algorithm. In the second step, genetic algorithm is adopted to select most important fuzzy rules in order to obtain compact fuzzy classification system. In the third step, a constraint real-coded genetic algorithm is used to optimize constructed compact fuzzy classification system to improve its precision.The performance of the proposed method is examined on Iris data and Wine data, and the simulation results show its validity. |