基于非负矩阵分解的半监督动态社团检测  

Semi-supervised dynamic community detection based on non-negative matrix factorization

作　　者：常振超[1] 陈鸿昶[1] 黄瑞阳[1] 于洪涛[1] 刘阳[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《通信学报》2016年第2期131-142,共12页Journal on Communications

基　　金：国家自然科学基金资助项目(No.61171108);国家重点基础研究发展计划基金资助项目(No.2012CB315901;No.2012CB315905);国家科技支撑计划基金资助项目(No.2014BAH30B01)~~

摘　　要：如何有效融合不同时刻的网络结构信息,是影响复杂网络中动态社团检测算法检测性能的关键和难点。基于此,提出了一种基于非负矩阵分解的半监督动态社团检测方法 SDCD-NMF,该方法首先有效提取了历史时刻所包含的稳定结构单元,然后将其作为正则化监督项,指导当前时刻的网络社团检测。在真实网络数据集上的实验表明,所提方法与已有方法相比具备更高的社团划分质量,更有利于探索网络的演变与发展规律。How to effectively combine the network structures on different time points was the key and difficulty to affect the performance of detection algorithms. Based on this, a semi-supervised dynamic community algorithm SDCD based on non-negative matrix factorization, which effectively extracted the historical stability structure unit firstly, and then use it as a regularization item supervision of nonnegative matrix decomposition, to guide the network community detection on current moment. Experiments on the real network data sets show that the method has a higher community detection quality compared with existing methods, which can accurately mine the relationship among different time, and explore network evolution and the law of development more advantageously.

关 键 词：半监督 动态 社团检测 非负矩阵分解 

分 类 号：TN915.0[电子电信—通信与信息系统]