基于混合隶属度随机块模型社会网络结构分析

发布时间：2018-05-30 07:43

本文选题：随机块模型 + 社会网络　；参考：《太原理工大学学报》2015年05期

【摘要】：由于大规模社会网络中存在着多种结构,且这些结构对于大规模社会网络的研究至关重要,但现有的结构发现方法大多只能够发现单一的结构或事先确定的结构,不能较为全面的反应大规模社会网络的特征。为解决上述社会网络中多结构发现的问题,引入了混合隶属度随机块模型MMSBM(Mixed Membership Stochastic Block Model)。它不仅能够生成不同结构的网络,同时还可以根据随机等价原则发现网络中的其他结构。通过在两个不同规模的微博数据集上进行结构发现实验,结果表明MMSBM能够同时发现社会网络中的多种结构,与实际观测结构基本吻合,但其计算复杂度较高,在实际应用中仍难以推广。
[Abstract]:Because there are many kinds of structures in large-scale social networks, and these structures are very important for the study of large-scale social networks, but most of the existing structure discovery methods can only find a single structure or a predetermined structure. It cannot fully reflect the characteristics of large scale social networks. In order to solve the problem of multi-structure discovery in social networks mentioned above, a mixed membership stochastic block model (MMSBM(Mixed Membership Stochastic Block Modeler) is introduced. It can not only generate networks with different structures, but also discover other structures in networks according to the principle of stochastic equivalence. The structure discovery experiments on two different size Weibo data sets show that MMSBM can simultaneously discover various structures in the social network, which is basically consistent with the actual observation structure, but its computational complexity is high. It is still difficult to popularize in practical application.
【作者单位】：太原理工大学计算机科学与技术学院;
【基金】：国家“863”高技术研究发展计划项目基金:基于用户兴趣模型的媒体大数据内容整合与可视化技术(2014AA015204) 山西省自然科学基金项目:动态社会网络隐结构推断与演化的关键技术研究(2014011022-1)
【分类号】：TP393.02

【参考文献】