基于支持向量机的模糊系统辨识研究  

Study on Fuzzy System Identification Based on SVM

作　　者：叶剑斌[1,2] 丁志燕[1,2] 

机构地区：[1]国网电力科学研究院,江苏南京210061 [2]南京南瑞集团公司,江苏南京210061

出　　处：《计算机与现代化》2011年第11期7-10,共4页Computer and Modernization

摘　　要：GK模糊聚类是一类广泛应用于分类的数据分析技术,能智能探测不同聚类的形状,但是存在迭代过程中聚类数恒定、公式中协方差矩阵要求非零等缺点。本文针对这些缺点,提出改进的聚类算法,针对现有的模糊辨识算法出现的维数灾难及函数逼近能力不高等问题,以语言模糊模型和缺少常数项的支持向量回归机的等价性分析为基础,提出一种支持向量机与模糊系统相结合的新辨识算法,并且利用梯度下降法对参数进行辨识;为了更好地缩减规则数及体现样本数据的信息,对输入的样本集又采用改进的GK模糊聚类对数据进行分类。GK fuzzy clustering is a data analysis technique that is widely applied to classification,and it can intelligently detect the shape of a different cluster.However,there are shortcomings about constant numbers of clusters in each iteration and the covariance matrix in the formula required to non-zero.In order to improve these shortcomings,this paper proposes a new fuzzy clustering algorithm.The existing fuzzy identification algorithm has the curse of dimensionality problem and it lacks of a solid theoretical foundation.The paper analyzes the equivalence of the lack of bias item support vector regression machine and language-based fuzzy model,proposes a new algorithm which combines support vector machines and fuzzy model identification.It also uses gradient descent method to identify parameters.In order to reduce the number of rules and reflect the data information,the paper adopts an improved GK fuzzy clustering for the input sample set to classify these data better.

关 键 词：支持向量机 模糊聚类 模糊系统辨识 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]