微博影响力传播模型的改进与验证

发布时间：2019-01-30 09:29

【摘要】：微博,作为交流最方便、信息传播最快捷的社交网络平台,已经深入人们的日常生活中。普通网民、公众人物、媒体、政府机构、企业等各个领域的人们越来越习惯通过微博来发布或传播最新消息。随着社交网络在研究领域的飞速发展,微博作为典型的社交网络,同样引起了学者们的广泛关注。本文从影响力传播入手,重点研究了微博中的影响力传播模型。微博中影响力传播主要通过信息传播来体现,因此分析信息传播的影响因素对改进影响力传播模型具有重要意义。信息传播的影响因素可以看作影响力传播模型中节点激活概率的影响因素。通过对现有模型的深入学习和对社交网络中信息传播的影响因素的分析,本文提出了三个微博中的节点激活概率影响因素——博文特征、用户间关系特征和用户特征,并给出了相应的计算方法,该方法称为TRU (Tweets, Relationships and Users)节点激活概率计算方法。其中博文特征主要度量纯文本、仅含URL链接的文本、仅含多媒体的文本和同时含URL、多媒体的文本四种形式的微博在传播中的影响程度。用户间关系特征从用户间行为惯性和用户间兴趣相似度展开,分析转发频度和兴趣相似度对信息传播的影响。用户特征主要通过用户影响力度量该用户在影响力传播中的作用。在独立级联模型中引入TRU节点激活概率计算方法并改进节点激活阈值计算方法的基础上,本文提出适用于微博的ICTRU (TRU measurement based on Independent Cascade model)影响力传播模型。通过新浪的微博开放平台提供的公开API,本文抓取了超过450万用户和其对应的微博信息来验证ICTRU模型的有效性。通过对比ICTRU模型的模拟传播图和真实微博的传播效果图,同时分析两者对应的转发树数据,证实了ICTRU模型可以有效的模拟微博中的影响力传播情况。同时对比ICTRU模型和原始的独立级联模型的传播情况,得出ICTRU模型可以更好的适用于微博中的影响力传播。
[Abstract]:Weibo, as the most convenient social network platform for communication and information dissemination, has penetrated into people's daily life. Ordinary Internet users, public figures, the media, government agencies, enterprises and other areas of people more and more accustomed to Weibo to publish or disseminate the latest news. With the rapid development of social networks in the field of research, Weibo, as a typical social network, has attracted extensive attention of scholars. This article starts with the influence dissemination, and focuses on Weibo's influence communication model. Weibo's influence communication is mainly reflected by information communication, so it is important to analyze the influencing factors of information communication to improve the influence communication model. The influencing factors of information transmission can be regarded as the influence factors of node activation probability in the influence propagation model. Based on the in-depth study of the existing models and the analysis of the influencing factors of information transmission in social networks, this paper puts forward three influential factors of the activation probability of nodes in Weibo, namely, the features of blog posts, the features of the relationship between users and the characteristics of users. The corresponding calculation method is given, which is called TRU (Tweets, Relationships and Users) node activation probability calculation method. The features of blog articles mainly measure the influence of Weibo in the transmission of pure text, including only the text of URL link, the text of multimedia and the text of URL, multimedia at the same time. The relationship between users is based on the behavior inertia of users and the interest similarity between users. The influence of forwarding frequency and interest similarity on information transmission is analyzed. User characteristics measure the role of the user in the dissemination of influence through the influence of the user. Based on the introduction of TRU node activation probability calculation method and the improvement of node activation threshold calculation method in the independent cascade model, a ICTRU (TRU measurement based on Independent Cascade model) influence propagation model for Weibo is proposed in this paper. Through the open API, provided by Weibo of Sina this paper fetches more than 4.5 million users and its corresponding Weibo information to verify the validity of the ICTRU model. By comparing the simulated propagation map of ICTRU model with that of real Weibo, and analyzing the corresponding forwarding tree data, it is proved that the ICTRU model can effectively simulate the influence transmission in Weibo. At the same time, by comparing the ICTRU model with the original independent cascade model, it is concluded that the ICTRU model is more suitable for the influence transmission in Weibo.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】