特征空间属性加权混合C均值模糊核聚类算法  

Mercer-kernel based mixed C-means fuzzy clustering algorithm with attributes weights in feature space

作　　者：贺杨成[1] 王士同[1] 江南[1] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122

出　　处：《计算机工程与应用》2011年第23期159-163,共5页Computer Engineering and Applications

基　　金：国家高技术研究发展计划(863)No.2007AA1Z158;国家自然科学基金重大研究计划(No.9082002);国家自然科学基金(No.60704047)~~

摘　　要：可能性聚类算法(PCM)通过引入可能隶属关系来提高聚类中心免于噪声干扰的能力,但是其往往趋向找到相同的集群。为了克服PCM算法的缺陷,PFCM算法同时利用隶属度与可能性把数据点划分到不同的集群中。提高了算法的抗噪能力。但PFCM算法对发现大小不相等的集群并不十分理想。因此提出了一种特征空间属性加权混合C均值模糊核聚类算法WKFM,该方法充分考虑了属性间的不平衡性,通过利用优化选取核参数的核函数把在原始空间中非线性可分的集群转化为高维空间中同质集群。实验结果表明,该算法能更好地发现含有噪音数据集的聚类中心,获得数据集质量更好的划分。A possibilistic clustering algorithm reduces the noisy influence on cluster centers by producing possibilities.However it often tends to find the identical cluster.To overcome this shortcoming,PFCM is proposed,which divides the data set into k different clusters through producing memberships and possibilities simultaneously,along with the cluster centers.But when two highly unequal sized clusters are given,PFCM fails to give the desired results.Therefore,a mercer-kernel based mixed C-means fuzzy clustering algorithm WKFM with attributes weights in feature space is proposed,which considers the imbalance between the attributes fully and uses kernel function to make it possible to cluster data that is linearly non-separable in the original space into homogeneous groups in the transformed high dimensional space with optimized kernel parameters. The experimental results show that the proposed algorithm can precisely find the ideal cluster centers and gives better results.

关 键 词：核 模糊聚类 模式识别 可能性聚类 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]