基于节点聚类系数的网络社区结构发现  

Discovery of Network Community Structure Based on Node Clustering Coefficient

作　　者：朱玲 余本国 冀庆斌[3] ZHU Ling;YU Ben-guo;JI Qing-bin(School of Science, North University of China, Taiyuan 030051, China;School of Medical Information, Hainan Medical College, Haikou 570216, China;School of Computer and Information Technology, Shanxi University, Taiyuan 030006, China)

机构地区：[1]中北大学理学院,山西太原030051 [2]海南医学院医学信息学院,海南海口570216 [3]山西大学计算机与信息技术学院,山西太原030006

出　　处：《中北大学学报（自然科学版）》2021年第5期435-440,459,共7页Journal of North University of China(Natural Science Edition)

基　　金：国家自然科学青年基金资助项目(61807005)。

摘　　要：研究了节点聚类系数与网络社区结构之间的关系.直接使用节点聚类系数不易刻画社区子图的高聚集特性,定义了一些基于节点聚类系数的社区度量,据此识别网络中的社区.首先,给出了基于聚类系数增大的社区间边判定规则,简称CCE规则;然后,利用CCE规则引出相似度矩阵,即网络密度矩阵;最后,通过网络密度矩阵来构造Laplacian矩阵,并进一步推导出通过计算Laplacian矩阵的特征值以及特征向量来实现社区结构划分的算法.三个真实网络数据的实验结果表明,算法不仅获得了令人满意的划分结果,而且还提高了算法的时间效率.The relationship between node clustering coefficient and network community structure was studied.It was not easy to characterize the high clustering characteristic of community subgraph by using node clustering coefficient directly.Based on this,some community measures based on the node clustering coefficient were defined to identify the communities in the network.First,based on the increase of clustering coefficient,an inter-community edge judgment rule,CCE rule,was given.Then,the similarity matrix,namely the network density matrix,was derived by using CCE rules.Finally,the Laplacian matrix was constructed by network density matrix,and the community structure partition algorithm based on the eigenvalues and eigenvectors of Laplacian matrix was derived.Experiments on real network data show that the proposed algorithm can not only obtain satisfactory segmentation results,but also improve the time efficiency of the algorithm.

关 键 词：网络 社区结构 节点聚类系数 网络密度矩阵 LAPLACIAN矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]