ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2014, Vol. 51 ›› Issue (10): 2225-2238.doi: 10.7544/issn1000-1239.2014.20131570

• 信息处理 • 上一篇    下一篇

TTRank:基于倾向性转变的用户影响力排序

段松青,吴斌,王柏   

  1. (北京邮电大学计算机学院 北京 100876) (dsq58629@163.com)
  • 出版日期: 2014-10-01

TTRank: User Influence Rank Based on Tendency Transformation

Duan Songqing, Wu Bin, Wang Bai   

  1. (School of Computer Science, Beijing University of Posts and Telecommunications, Beijing 100876)
  • Online: 2014-10-01

摘要: 近年来,不少学者从回复关系的角度分析用户影响力,但存在回复关系稀少、帖子内容被忽视、不能动态更新等问题.为弥补这些不足,提出了一种基于倾向性转变的用户影响力分析方法.先计算帖子的影响力,再提出“局部回复链”的概念,引入间接回复关系计算方法,增加了帖子之间的回复关系;然后对局部回复链,分析用户倾向性变化的过程,得到用户影响他人和受影响的程度,最终获得用户在指定范围内的影响力排名.该算法与10种经典的影响力分析算法对比以及实例分析的结果,说明该算法能从其他角度更好地刻画用户形象.

关键词: 社会化媒体, 论坛, 影响力分析, 用户影响力, 倾向性转变

Abstract: In recent years, many scholars have studied the problem of how to analyze user influence from the perspective of reply relationships. But there are various problems, such as reply relations are scarce, the content of posts are ignored, don’t support dynamic updates and so on. To compensate for these deficiencies, a user influence analysis method based on tendency transformation is proposed. Firstly, the post influence is calculated, the concept of “local reply chain” (LRC) is proposed, and the calculation method of reply indirect relationships is introduced, so the number of reply relationship is drastically increased. Secondly, the process of user tendency transformation is analyzed based on LRC, and the degrees of user impacting on others and being affected by others are gotten. Finally, the user influence ranking within the specified range is obtained. Compared with 10 kinds of classical influence analysis algorithms and used to analyze realistic data, the algorithm can characterize the user image better from the other point of view.

Key words: social media, forum, influence analysis, user influence, tendency transformation

中图分类号: