高级检索
    张 玥 张宏莉 张伟哲 卢珺珈. 识别网络论坛中有影响力用户[J]. 计算机研究与发展, 2013, 50(10): 2195-2205.
    引用本文: 张 玥 张宏莉 张伟哲 卢珺珈. 识别网络论坛中有影响力用户[J]. 计算机研究与发展, 2013, 50(10): 2195-2205.
    Zhang Yue, Zhang Hongli, Zhang Weizhe, and Lu Junjia. Identifying the Influential Users in Network Forum[J]. Journal of Computer Research and Development, 2013, 50(10): 2195-2205.
    Citation: Zhang Yue, Zhang Hongli, Zhang Weizhe, and Lu Junjia. Identifying the Influential Users in Network Forum[J]. Journal of Computer Research and Development, 2013, 50(10): 2195-2205.

    识别网络论坛中有影响力用户

    Identifying the Influential Users in Network Forum

    • 摘要: 网络论坛已经成为网络用户发布信息的重要渠道.在论坛中对热点话题的讨论影响着物理世界中人们的看法、观点以及国家政策法规的制定.由此提出一系列研究问题:如何计算用户影响力?不同主题不同时间下用户影响力如何比较?用户影响力发展趋势如何?根据幂律规律,大量用户形成“长尾”,如何识别有影响力用户?以主题为单位,提取用户间回复关系,构建用户对话关联图,回复次数和回复长度形成用户行为特征,入度和出度形成网络结构特征.在Pagerank算法基础上,结合用户行为特征以及用户间关联网络特征,提出基于多属性的用户影响力排序算法(multiple attributes rank, MAR).并依据发表时间进行时间段切分,得到论坛上每日有影响力用户排行榜,进一步分析了有影响力用户演化趋势.以天涯网络论坛真实数据进行实验,从多角度评价有影响力用户以及MAR排序算法,得到一些有趣结论并对未来工作进行了展望.

       

      Abstract: Network BBS has been a very important way for Web user to publish information. The discussion on hot topics of BBS affects people's view, attitude to physical world, and even enactment. So a series of problems are put forward: How to compute the network user's influence? What is the user's influence trend? How to compare user's influence in different topics? By the power law, a lot of users are the “long tail”. How to identify the influential users? Taking theme as unit, extracting the reply information among users, we build user cascade relation graph, in which the number of reply and the length of reply make up user's behavior features. In-degree and out-degree make up structural features. Based on the Pagerank algorithm, introducing user's behavior feature and the user relational networks, this paper designs a multiple attributes rank (MAR) algorithm. Influential user's evolution trend is analyzed by time span division, based on user's interest difference on topic at different time. To illustrate these problems, we have conducted experiments with data from Tianya BBS, and evaluated multi-facets of issues of identifying influential users and MAR algorithm. We have summarized some interesting findings and expec ted the future work.

       

    /

    返回文章
    返回