摘要
针对如何在高校论坛中识别出高影响力用户的问题,提出了一种结合时间因子的论坛用户影响力计算方法。在PageRank算法的基础上,通过分析论坛用户回帖行为随时间变化的规律引入用户影响力的衰减系数,并根据用户间的交互情况,通过用户间的交互次数和用户主题参与度来优化用户影响力的分配形式。实验结果表明:在相同时间段内,改进后的算法相对于PageRank算法能够识别出具有更强中心性特性和更高用户覆盖率的高影响力用户。在连续时间段上,通过引入衰减系数能够发掘当前真正具有影响力的用户并保持稳定的用户覆盖率。
In order to identify high-impact users in the campus forums,an algorithm combined with time factor is proposed.Based on PageRank algorithm,the attenuation coefficient of user' s influence is introduced by analyzing the change law of user's reply behavior,and the propotion of user' s influence assigned to others is determined by the number of interactions between users and user' s subject participation. Experimental results indicate that the improved algorithm can identify the influential users with higher centrality and user coverage ratio than PageRank algorithm in the same time period. In the continuous period of time,by introducing the attenuation coefficient,the real high-impact users can be identified and stable user coverage ratio can be maintained.
出处
《智能计算机与应用》
2017年第5期13-16,共4页
Intelligent Computer and Applications
基金
国家重点研发计划资助项目(2016YFB0801200)
关键词
影响力
时间因子
衰减系数
主题参与度
中心性特性
覆盖率
influence
time factor
attenuation coefficient
subject participation
centrality
coverage ratio