基于相似度聚类的二分网络社区发现算法  

The bipartite-network community detection algorithm based on the similarity clustering

作　　者：张晓琴[1] 刘莉楠 ZHANG Xiao-qin;LIU Li-nan(School of Mathematical Sciences,Shanxi University,Taiyuan 030006,Chin)

机构地区：[1]山西大学数学科学学院,山西太原030006

出　　处：《云南民族大学学报（自然科学版）》2018年第4期307-314,共8页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(61573229);山西省回国留学人员科研资助项目(2017-020);山西省基础研究计划项目(201701D121004);山西省高等学校教学改革创新项目(J2017002)

摘　　要：针对往往不能提前预知社区个数的情况,提出了基于相似度聚类的二分网络社区发现算法(similarity clustering algorithm,简称SCA).算法通过计算U类节点之间的相似度获得核心节点,同时选取核心节点邻域中的节点扩展得到社区,将未划分到社区中的孤立点和只包含一个节点的社区分别放入与之联系最紧密的社区中,最后V类节点划分到已有的社区中得到完整的社区划分结果.通过在人工数据集与真实网络上的分析,分别利用归一化互信息和模块度作为评价指标,实验结果表明,SCA比BRIM等算法能够更有效挖掘二分网络社区结构,具有比较良好的社区划分效果.Aiming at the problem concerning the determination of the number of communities in advance,this paper proposes a community detection algorithm,that is,the Similarity Clustering Algorithm（ SCA） in the bipartite networks. The algorithm obtains the core nodes by calculating the similarity between the U-type nodes. Meanwhile,through selecting the nodes in the core node＇s neighborhood to expand to get the community and the outliers that are not classified into the community,and some communities that contain only one node are placed in the most closely connected communities. Finally,V-type nodes are divided into the existing communities to obtain complete community-division results. Through the analysis on artificial networks and real-world networks,the normalized mutual information and modularity are used respectively as evaluation indicators. The experimental results show that SCA is able to mine the bipartite-network community structure more effectively and has a better community division than the BRIM and other algorithms.

关 键 词：二分网络 社区发现 相似度 归一化互信息 模块度 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]