基于簇内不平衡度量的粗糙__-means 聚类算法.pdfVIP

第28 卷 第10 期 控 制 与 决 策 2013 年 10 月 Vol. 28 No. 10 Control and Decision Oct. 2013 文章编号: 1001-0920 (2013) 10-1479-06 基于簇内不平衡度量的粗糙-means 聚类算法 a a b 张腾飞 , 陈 龙 , 李 云 (南京邮电大学a. 自动化学院，b. 计算机技术研究所，南京210023) 摘 要: 粗糙-means 聚类算法是一种有效的处理聚类边界模糊问题的算法, 但大多数算法对簇的下近似集和边界 中的对象使用统一的权值, 忽略了簇内对象之间的差异性. 针对这一问题提出一种新的改进算法, 通过对簇内的每个 对象加入簇内不平衡度量, 以区分不同对象对簇的贡献程度, 使得聚类结果簇内更紧凑、簇间更疏远. 不同数据集的 仿真实验结果表明, 所提出算法可以有效提高聚类结果的精度. 关键词: 簇内不平衡度量；粗糙集；粗糙-means 聚类 中图分类号: TP18 文献标志码: A Rough -means clustering based on unbalanced degree of cluster a a b ZHANG Teng-fei , CHEN Long , LI Yun (a. College of Automation，b. Institute of Computer Technology ，Nanjing University of Posts and Telecomunications, Nanjing 210023 ，China ．Correspondent ：ZHANG Teng-fei ，E-mail ：tfzhang@126.com) Abstract: Rough -means clustering is a valid algorithm to process the inseparability of border of clusters. But to most algorithms, weights of objects in the lower approximate set or the upper approximate set are all the same without paying attention to the diversity in clusters. Therefore, a new algorithm is proposed. The algorithm can make the cluster has a more compact center, and the borders are separated each other with the unbalanced degree of cluster which means the contribution of an object to the cluster. The simulation analysis shows that this algorithm can improve the precision of the clustering results effectively. Key words: unbalanced degree of cluster ；rough set ；rough -means clustering 0 引引引 言言言 聚类是一种将数据对象分成类或簇的过程, 它要 集, 每一个对象或者确定属于某个簇的下近似集, 或 求簇内数据对象高度相似, 而簇间数据对象高度相异. 者同时属于多个簇的上近似集. 该算法虽然