模糊系统的递阶遗传算法设计新方法

New Method for Design of Fuzzy Systems Based on Hierarchical Genetic Algorithm

针对模糊系统的特点和根据输入-输出数据,应用递阶遗传算法设计模糊系统。现有的模糊系统设计方法大多只能训练模糊系统的模糊集的中心参数、高斯隶属度函数的系数、中心参数和扩展参数,系统的模糊规则得预先用某种方法确定。利用很好设计的递阶遗传算法能够把模糊系统的模糊规则数目和参数同时通过训练确定。通过对混沌时间序列进行预测仿真,结果证明用该方法设计的模糊系统预测的精确度是令人满意的,文中提出的方法是可行的。

A hierarchical genetic algorithm was proposed to train parameters of fuzzy systems, such as centers of fuzzy set, coefficients, centers and spreads of Gaussian membership function. In addition, the number of fuzzy rules was determined at the same time during training, Training and test based on practical data sets were carried out respectively, and a good performance of the new algorithm was demonstrated. The model was used to forecast the chaotic time series. It is shown that the model based on hierarchical genetic algorithm proposed is simple and effective.