微博网络中社区意见领袖发现技术研究

发布时间：2019-06-26 11:44

【摘要】：微博作为近些年新兴的社交网站,对整个互联网产生了巨大影响,微博服务极大的拉近了人与人之间的距离,让人们拥有了新的合作与交流方式。对于微博网络中的用户,每一个用户都是信息传播的入口,这种新的方式可以爆发式的产生大量的实时消息。因此,用户之间的关注关系与转发关系构成了一个庞大且复杂的微博网络。在微博网络逐渐形成的过程中,微博社区也逐渐形成并且达到一定规模。这是一种普遍的社会网络现象,一般这种社区的内部具有很强的相关性与联系性。如果能够准确的划分出各个社区,这对于进一步改善微博服务质量是一个前提条件。通过社区发现算法的比较,本文研究在大规模网络中标签传播算法(Label Propagation Algorithm,简称LPA)是一个快速有效的社区发现方法,也是目前应用范围比较广的一个算法。但是,标签传播中的随机节点访问降低了算法的鲁棒性,对于识别社区结构的稳定性也造成了一定的影响。因此本文提出一种具有节点优先度的标签传播算法(Precedential Label Propagation Algorithm,简称PLPA),通过引入节点访问优先度的机制来改善随机节点访问带来的鲁棒性下降的缺陷。并通过对比实验验证算法的有效性与稳定性。另一方面,本文针对微博网络中的用户排名进行了系统地研究,而大部分的学者在进行该方面研究时只考虑单一的用户关注关系,或者用户发布微博的数目,而在微博中有着较高影响力的用户通常会发布一些价值含量较高的信息,并且有着大量的跟随用户。本文提出一种基于时效性以及随机游走策略的用户排名算法(本文称为User-Weibo Rank,简称UWRank),在该算法所强调的用户排名策略中,同时考虑用户的关注关系与微博发布影响力这两大因素。针对微博信息的高实时性特点,融合微博内容的时效性作为衡量影响力的因素。并且通过实验,验证算法再保持较好结果的前提下,算法能够达到较好的收敛效果。
[Abstract]:As a new social network site in recent years, the micro-blog has a great impact on the whole Internet, and the micro-blog service has greatly shortened the distance between people and people, so that people have a new way of cooperation and communication. For users in the microblogging network, each user is an entry for information dissemination, and this new way can generate a large number of real-time messages. Therefore, the relationship between the users and the forwarding relationship form a large and complex micro-blog network. In the process of the gradual formation of the micro-blog network, the micro-blog community is gradually formed and reached a certain scale. This is a common social network phenomenon, and generally the interior of such a community has a strong correlation and contact. This is a prerequisite for further improving the quality of the microblogging service if the communities can be accurately divided. In this paper, the label propagation algorithm (LPA) in a large-scale network is a fast and effective method for community discovery and a wide range of algorithms. However, random node access in the label propagation reduces the robustness of the algorithm and has a certain effect on the identification of the stability of the community structure. This paper proposes a label propagation algorithm (PLPA) with node priority, which can improve the robustness of random node access by introducing the mechanism of node access priority. And the validity and the stability of the algorithm are compared by comparing the experimental verification algorithm. On the other hand, this paper makes a systematic study of the user's ranking in the micro-blog network, and most of the scholars consider only a single user-interest relationship when doing the research, or the number of users to publish the micro-blog, Users with high influence in the system often release high-value information and have a large number of follow-up users. This paper presents a user-ranking algorithm based on time-effectiveness and random walk strategy (referred to herein as" User-Weibo Rank "), in which the user's attention relationship and the influence of microblogging are considered in the user's ranking strategy emphasized by the algorithm. Aiming at the high real-time character of the micro-blog information, the time-effectiveness of the content of the micro-blog is used as a factor to measure the influence. And the algorithm can achieve better convergence effect on the premise of the experiment and the verification algorithm to keep the better result.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】