改进的基于二次型模糊c均值聚类模型  

Improved fuzzy c-means model based on quadratic

作　　者：陈加顺[1,2] 皮德常[1] 

机构地区：[1]南京航空航天大学计算机科学技术学院,江苏南京210016 [2]淮海工学院计算机工程学院,江苏连云港222003

出　　处：《系统工程与电子技术》2013年第7期1547-1553,共7页Systems Engineering and Electronics

基　　金：国家高技术研究发展计划(863计划)(2007AA01Z404);江苏省普通高校研究生科研创新计划(CXLX11_0206)资助课题

摘　　要：针对模糊聚类算法对点数据集聚类的敏感性以及区间类型数据聚类效果不明显等问题,提出了基于二次型距离改进的模糊可能性c均值(fuzzy-possibilistic c-means,FPCM)聚类算法。首先分析了区间数据的特征,引入了区间值的数学表示方法,在此基础上提出了3种不同的基于区间数据距离度量方法以及相应权重矩阵的计算方法,通过建立拉格朗日方程对目标方程优化,求得聚类中心、隶属度以及可能性迭代方程,并证明目标方程的收敛性,最后给出了算法执行步骤。在不同类型的数据集上实验,证明算法在点数据集和区间数据集上都具有较好聚类性能.Aiming at the problems of most fuzzy clustering algorithms being sensitive to point data sets and the shortcoming of unobvious clustering results and so on,an improved fuzzy possibilistic c-means（FPCM） clustering algorithm based on quadratic distance is proposed.Firstly,the feature of interval-valued data is analyzed and the mathematical representation method of interval-valued sample data is introduced.On the basis of these,three measure methods between interval-valued sample data and prototypes and the corresponding computing methods of weight matrices are presented,and an objective function is proposed.The iterative function of centroid,membership and typicality is acquired by constructing a Lagrange equation,and the conrergence of the iterative function is proved.Finally,the steps of the proposed algorithm are provided.Experiments on different types of three data sets show that the proposed algorithm has a good performance not only on point prototype but also on interval-valued prototype.

关 键 词：模糊聚类 改进模糊可能性c均值 二次型距离 权重矩阵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]