基于联系度的主题关注网络社区发现方法研究

Community discovering based on connection degree in topic-attention network

目前,以兴趣或主题分享等为目的的兴趣型社交网络则引领着社交网络改革的浪潮。融合社交关系和兴趣爱好关系构建一个新型社交网络模型——主题关注模型。在此模型基础上,采用集对联系度刻画顶点间相似性度量指标,该度量方法可以更好地刻画网络结构特征,提高传统局部相似性度量指标对某些顶点间相似性值的计算精度,降低全局相似性度量指标的计算复杂度。综合考虑主题影响和社交关系,将集对联系度与凝聚型聚类算法相结合,提出一种新的主题社区发现方法。在Karate网络和豆瓣数据集上进行主题社区发现,实验结果表明,考虑主题影响的划分具有更好的社区结构。

At present,the interest’s social networks,which share interests or topics,lead to a reform wave of social network.A new social network model,the topic-attention model,fusing the social relations and the topic-attention relation is constructed.Based on this model,firstly,using set pair connection degree to define the similarity between vertices,the measurement can better describe network structure characteristics,overcome the under-estimating for some similarity between vertices based on traditional local structures,and reduce the computational complexity of global similarity indices.Secondly,considering the influence of the topic and the social relationships,a new method of community discovery is proposed based on the set pair connection degree and the clustering algorithm.Finally,in Karate network and douban data set,experiment on the topic community mining,the results show,considering the impact of the topic influence has a better community structure.