在线社会网络用户影响力算法研究

发布时间：2018-06-21 02:09

本文选题：社会网络分析 + 数据挖掘　；参考：《天津大学》2014年硕士论文

【摘要】：随着在线社会网络应用的发展,在线社会网络分析方法受到研究人员的关注。同时,社交网站产生海量关于用户朋友关系以及自身行为(例如用户观点、评论和照片等等)的数据。现今,用户通过网络进行交互的行为基本都发生在社交网络。通过社交网站,人类可以更加方便和高效的同他人进行交流。所有看似微小的信息,经过社交网站数以百万人的行为操作,都能构成一个巨大并且快速迭代的数据集。面对如此丰富的信息,学术界和工业界开始进行相关研究。搜索、关联性研究以及对于个人用户数以百万计的发布信息的研究,成为一个重要的研究方向。然而,现有的行为分析方法大多面向全局网络,以某用户兴趣为中心的局部网络用户行为算法研究比较缺乏;现有方法主要考虑行为之间的相关作用,忽略亲密度、影响力和用户交互行为等社交因素。论文基于以上问题,提出基于用户亲密度和影响力的社交圈用户影响力挖掘算法。基于海量社会网络数据,提出基于PageRank挖掘社会网络关键用户的PageRank-Based(PRB)排序算法。该算法可适用于现实世界中产生的大量数据。在每个阶段,论文首先对问题本身进行一个简短的陈述,然后通过基于PageRank的关键节点挖掘算法,实现对大数据用户行为的分析。为验证PRB排序算法的有效性,论文提出基于聚类系数的节点重要度系数指标,进行PRB算法有效性验证。然后,将PRB排序算法得到的关键用户节点运用于D-S证据理论进行用户行为分析。最后,论文提出基于关键节点的局部社区发现算法,将PRB排序算法得出的关键用户节点作为潜在种子节点,通过计算节点间的亲密度来剔除可能出现的重叠社区。实验结果证明所提算法的正确性和有效性。
[Abstract]:With the development of online social network application, online social network analysis method has attracted much attention. At the same time, social networking sites produce huge amounts of data about user friendships and their own behavior (such as user views, comments, photos, etc.). Nowadays, the interaction behavior of users through the network basically occurs in the social network. Through social networking sites, humans can more easily and efficiently communicate with others. All the seemingly minuscule information, manipulated by millions of people on social networking sites, can make up a large and fast iterative data set. In the face of such a wealth of information, academia and industry began to carry out related research. Search, relevance research and the study of millions of individual users to publish information, become an important research direction. However, most of the existing behavior analysis methods are oriented to the global network, and the research on the local network user behavior algorithms based on a user's interest is relatively lacking, and the existing methods mainly consider the correlation between behaviors and ignore the affinity. Social factors such as influence and user interaction. Based on the above problems, this paper proposes a social circle user influence mining algorithm based on user affinity and influence. Based on massive social network data, a PageRank-Based PRB-based sorting algorithm based on PageRank mining key users of social networks is proposed. The algorithm can be applied to large amounts of data generated in the real world. In each stage, the paper first makes a brief statement of the problem itself, and then realizes the analysis of the behavior of big data users by mining the key nodes based on PageRank. In order to verify the validity of PRB sorting algorithm, this paper proposes a node importance coefficient index based on clustering coefficient, and verifies the validity of PRB algorithm. Then, the key user nodes obtained by PRB sorting algorithm are applied to D-S evidence theory for user behavior analysis. Finally, a local community discovery algorithm based on key nodes is proposed. The key user nodes obtained by PRB sorting algorithm are regarded as potential seed nodes, and the overlapping communities may be eliminated by computing the affinity between nodes. Experimental results show that the proposed algorithm is correct and effective.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【相似文献】