融合自组织映射与Wang-Mendel方法的模糊规则提取  

Fuzzy Rule Extraction by Fusing SOM and Wang-Mendel Method

作　　者：於东军[1] 谌贻华[1] 于海瑛[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,江苏南京210094

出　　处：《南京理工大学学报》2011年第6期759-763,共5页Journal of Nanjing University of Science and Technology

基　　金：江苏省自然科学基金(BK2011371);江苏高校优势学科建设工程;南京理工大学自主科研专项计划(2011YDXM19)

摘　　要：为解决经典Wang-Mendel方法提取模糊规则时得到的规则数目较多的问题,该文提出一个融合自组织映射(SOM)网络和Wang-Mendel方法的规则提取方案。利用SOM对原始样本进行学习,再使用Wang-Mendel方法从SOM的原型向量中提取规则。因为SOM的原型向量学习得到了原始样本的模式分布规律,并且原型向量的数量远小于原始学习样本的数量,所以该方法降低了提取的规则数目及耗费时间。有关混沌时间序列预测的实验结果表明了该文方法的有效性。To solve the problem that too many fuzzy rules may be extracted by the classic Wang-Mendel method,a novel rule extraction method which combines self-organizing map（SOM）and Wang-Mendel method is presented.Original samples are firstly learned by the SOM,and rules are further extracted from the codebook vectors of the learned SOM by the Wang-Mendel method.The proposed method decreases the number of the extracted rules and the time consumption because the codebook vectors of the learned SOM achieves the pattern distribution knowledge of the original samples and the number of the codebook vectors is far less than that of the original samples.Experimental results on chaotic time series prediction demonstrate the effectiveness of the proposed method.

关 键 词：自组织映射 模糊规则提取 混沌时间序列 模式分布 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]