面向多源社交网络的社团结构特征研究

发布时间：2018-05-17 02:46

本文选题：社交网络 + 社团发现　；参考：《工程科学与技术》2017年S2期

【摘要】：为了研究社交网络社团结构对舆情传播的影响,本文对比分析了多源社交网络的社团结构特性及传播特性,并且利用COPRA算法和LFM算法进行了社交网络重叠社团研究,提出一种基于节点度过滤的LFM改进方法——NF-LFM算法。该算法先对好友关系网络中节点度小于某一阈值的节点进行过滤,再对剩下的好友关系网络进行社团划分。研究发现:1)人人网、QQ空间、新浪微博都具有明显的社团结构特性,其中,人人网和QQ空间的社团结构特性强于新浪微博;2)在不考虑社交网络用户活跃度的情况下,舆情信息在人人网上扩散范围最广,新浪微博次之,QQ空间扩散较慢。本文提出的改进方法能解决现有算法社团划分结果分辨率低的问题,且有效弥补了LFM算法在大规模社团发现时陷入无限迭代过程而导致时间复杂度高的缺点,将其应用于经典数据集中也符合理论预期结果。本文的研究结果将有助于进一步理解和认识社交网络社团结构对舆情传播的影响,同时对于网络群体事件发现和舆情监控及引导等具有重要意义。
[Abstract]:In order to study the influence of social network community structure on public opinion transmission, this paper compares and analyzes the community structure and propagation characteristics of multi-source social network, and uses COPRA algorithm and LFM algorithm to study the overlapping community of social network. This paper presents an improved LFM algorithm based on nodal filtering-NF-LFM algorithm. The algorithm first filters the nodes whose degree of nodes are less than a certain threshold in a friend network, and then divides the remaining friend networks into groups. The study found that: 1) Renren / QQ space and Sina Weibo have obvious community structure characteristics, among which Renren and QZone have stronger community structure than Sina Weibo 2) without considering the activity of social network users. Public opinion on Renren spread the widest range, Sina Weibo followed by QQ space diffusion slow. The improved method proposed in this paper can solve the problem of low resolution of community partition results of existing algorithms, and effectively make up for the high time complexity caused by the LFM algorithm falling into infinite iterative process in large-scale community discovery. Its application in classical data sets is also in line with the expected theoretical results. The results of this paper will be helpful to further understand and understand the influence of the social network community structure on the dissemination of public opinion, and it is of great significance for the discovery of network group events and the monitoring and guidance of public opinion.
【作者单位】：四川大学计算机(软件)学院;四川大学网络空间安全研究院;四川大学网络空间安全学院;
【基金】：国家科技支撑计划资助项目(2012BAH18B05) 国家自然科学基金资助项目(61272447) 四川大学青年教师启动基金资助项目(2015SCU11079)
【分类号】：TP393.09

【参考文献】