快速聚类和统计信息优化准则在模糊建模中的应用

The Application of Fast Clustering and Statistic Information Optimal Criterion to Fuzzy Modeling

给出了一种快速聚类方法,得到采样数据的聚类中心,用高斯隶属函数获得T- S模型的前提隶属度,然后采用正交最小二乘和“客观”的统计信息准则来选择一些重要的模糊规则,简化模糊模型,提高辨识精度和泛化能力,奇异值分解方法得到结论参数。最后通过仿真实例验证了此方法的有效性。

A fast clustering method is presented, which can get cluster centering of sample data. The antecedent membership degree of T-S model is obtained by Gaussian membership function, then some important rules using orthogonal least-square and 'objective' statistical information criterion are selected to reduce fuzzy model, and improve the pricision and genereli-zation ability of the fuzzy model. And the singular value decomposition is used to get consequent parameters. Finally, the effectiveness of this method is demonstrated by simulation. ©, 2005, Science Press. All right reserved.