基于链接相似性聚类的重叠社区识别  

Overlapping Community Detection Based on Link Similarity Clustering

作　　者：张桂杰[1,2] 张健沛[1] 杨静[1] 辛宇[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001 [2]吉林师范大学计算机科学与技术学院,吉林四平136000

出　　处：《电子学报》2015年第7期1329-1335,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61370083;No.61073043;No.61073041);高等学校博士学科点专项科研基金(No.20112304110011;No.20122304110012);吉林省教育厅"十二五"科学技术研究基金(No.吉教科合字2013[207])

摘　　要：社区结构是社会网络最普遍和重要的拓扑属性之一,提出一种基于链接相似性聚类的重叠社区识别算法.该算法首先根据相邻链接的度分布状态,提出链接间的相似性度量方法;其次以链接相似性矩阵为输入,以链接社区的最优划分为目标,建立链接局部相似性聚类算法,实现了重叠社区的有效识别;然后对链接社区进行优化,解决了可能出现的过度重叠及孤立社区问题;最后在真实网络及人工合成网络上的实验验证了算法的高效性.Community structure is one of the most common and important social network topological properties. This paper proposes a link community detection algorithm based on hierarchical clustering. Firstly,the algorithm sets up similarity measure according to the degree distribution of links nearby; then sets up local link similarity clustering algorithm which takes the similarity matrix as input with the purpose of detecting the best link community; further more realizes link community detection effectively. And then,optimize the link community to solve the problem of excessive overlapping and isolated community. Experiment results based on real world and computer generated networks showthat the algorithm is highly efficient.

关 键 词：社区识别 链接社区 局部链接相似性度量 层次聚类 重叠社区 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]