基于GMDH模型及模糊聚类的特征提取研究  

Research on Fuzzy Clustering-based GMDH Model to Feature Selection

作　　者：赵珩君[1] 刘斌[1] 腾格尔[2] 林亮亮[1] 肖进[2] ZHAO Hengjun;LIU Bing;TENG geer;LIN Liangliang;XIAO Jin(School of economics and management, Sichuan Radio and TV University, Chengdu 610073;School of Business, Sichuan University, Chengdu 610064)

机构地区：[1]四川广播电视大学经济管理学院,成都610073 [2]四川大学商学院,成都610064

出　　处：《科学与管理》2016年第6期69-76,共8页Science and Management

基　　金：国家自然科学基金委面上项目(71471124);成都市科技局软科学项目(2015-RK00-00259-ZF);四川广播电视大学重点教改项目(XMZSXX2016003Z)

摘　　要：特征提取算法可以去除目标数据中的冗余特征、无关特征甚至噪声特征,从而得到一个无冗余、无噪声的样本集,有助于提高目标对象的识别率以及数据的挖掘速度。现有的特征提取方法在定性数据及噪声数据的处理上存在局限性,而定性数据及带噪声数据在现实建模过程中是不可避免的。本文从特征提取需解决的根本问题出发,就如何确定特征子集并选择适当的隶属函数来表示模糊子空间,使模糊规则归纳模型有最大的识别率及抗干扰性的方法进行讨论、研究。Feature extraction algorithm can remove redundant features, irrelevant features and even noise characteristics of thetarget data, and thus obtain a non-redundant and non-noisy sample set. It helps to improve the recognition rate of the target objectand the data mining speed. The existing feature extraction methods have limitations in the process of qualitative data and noise data.However, the qualitative data and the noisy data are inevitable in the process of real modeling. Based on the fundamental problemof feature extraction, this paper sets out how to determine the feature subset and select the appropriate membership function torepresent the fuzzy subspace. The method of fuzzy rule induction model which has the maximum recognition rate and anti-jamming isdiscussed.

关 键 词：特征提取 模糊聚类 隶属函数 

分 类 号：O159[理学—数学] TP311[理学—基础数学]