摘要
提出了一个用于微博网络社区发现的模型WB-MMSB,该模型考虑了微博网络中节点存在的单向关系,节点的社区隶属度从链入主题隶属度和链出主题隶属度两个方面表示。用指数族分布和平均场变分推理方法推导了模型中各变量的表示,并用SVI算法计算模型涉及的参数。实验在新浪微博数据集上进行,采用归一化互信息和困惑度进行评估,结果表明,WB-MMSB模型的社区发现能力优于aMMSB模型,并且其收敛速度快于aMMSB模型。
Considering the nodes of Mico-blog network have single direction relations,a new model WB-MMSB was put forward for community detection,which uses directed edges to embody the direction relations of nodes,and two aspects link-in and link-out are used to quantify the community membership of nodes.Exponential family distribution and meanfield variational inference method were used to inference the representations of variables in this model,and SVI algorithm was used to compute relating parameters.Experiments adopted Sina-Weibo dataset and NMI to testify the performance of WB-MMSB.The results indicate that the community detection ability of WB-MMSB model is better than aMMSB model,and the convergence rate of WB-MMSB model is faster than aMMSB model.
出处
《计算机科学》
CSCD
北大核心
2015年第3期65-70,共6页
Computer Science
基金
中国博士后科学基金项目(20070420700)
河北省自然科学基金项目(F2011201146)资助
关键词
微博网络
社区发现
混合隶属度随机块模型
重叠社区
Micro-blog network
Community detection
Mixed membership stochastic block model
Overlapping commu-nities
