文献类型：期刊文献

中文题名：一种新型的模糊C均值聚类初始化方法

英文题名：A Novel Initialization Method for Fuzzy C-means Algorithm

作者：刘笛[1];朱学峰[2];苏彩红[2]

机构：[1]华东理工大学自动化系,上海200237;[2]华南理工大学自动化学院,广东广州510640

年份：2004

卷号：21

期号：11

起止页码：148

中文期刊名：计算机仿真

外文期刊名：Computer Simulation

收录：CSTPCD;;CSCD:【CSCD_E2011_2012】；

基金：广东省科技攻关项目 ( 2KM0 0 60 8G)

语种：中文

中文关键词：模糊C均值聚类;初始聚类中心;不完全匹配;免疫记忆

外文关键词：FCM; Initial cluster centers; Incomplete matching; Immunological memory

摘要：模糊C均值聚类 (FCM)是一种广泛采用的动态聚类方法 ,其聚类效果往往受初始聚类中心的影响。受自适应免疫系统对入侵机体的抗原产生免疫记忆的机理启示 ,提出了一种新的产生初始聚类中心的方法。算法中 ,待分析的数据被视为入侵性抗原 ,产生的记忆细胞作为聚类分析的初始中心。克隆选择用来产生抗原的记忆细胞群体 ,免疫网络理论则用来抑制该群体规模的快速增长。实验结果表明免疫记忆机理用于FCM初始中心的选择是可行的 ,不仅提高了FCM算法的收敛速度 。
The fuzzy C-means algorithm (FCM) is widely used for dynamic clustering. The performance of FCM depends on the selection of the initial cluster center. Inspired by the mechanism that the adaptive immune system remembers the antigen exposed to the body before, a novel algorithm is proposed for the generation of the initial cluster center. In this algorithm, the data set to be analyzed is taken as the invading antigen and the memory cell generated acts as the initial cluster center. While the clonal selection principle is responsible for generating the memory cell population, the immune network theory prevents the population size from increasing quickly. The experimental results have shown the feasibility of applying the immunological memory mechanism to the selection of initial centers in dynamic clustering. By adopting this algorithm, not only the accuracy and the convergence speed of FCM are improved, but also the number of clusters does not require to be predefined; it depends on the threshold to be chosen.