A Novel Genetic-fuzzy Modeling Method for Complex Systems
Shen Yuming · Acta Simulata Systematica Sinica · 2003
In this paper, we present a novel genetic-fuzzy modeling method for complex systems. Firstly, the input space is clustered by competitive learning algorithm. With the result, we get the membership function of the antecedent fuzzy sets. The consequent parameters of each individual rule are obtained as a local least squares estimate. Thus, the coarse T-S fuzzy model is taken. Secondly, the parameters of rules are encoded and the fuzzy systems are optimized by real-coded genetic algorithm. At last, digital simulation is performed to demonstrate the validity of our approach.