人类动力学视角下在线社区信息交流行为研究

Research on Information Communication Behavior in Online Community Based on Human Dynamics

【目的/意义】从人类动力学视角探索在线社区信息交流行为的独特规律,为信息交流行为定量建模与在线社区服务优化提供理论依据。【方法/过程】通过应用、改进人类动力学的经典分析指标,以天涯社区中79501条主贴及其关联的979718条回贴为源数据,从时间间隔、阵发性、记忆性、波动性、有序性等维度,分别针对个体用户回帖行为、单条主贴下群体用户回帖行为、单条主贴下群体回帖行为涌现(热门话题)进行统计分析。【结果/结论】在个体层面,绝大多数用户回贴行为的时间间隔分布服从幂律分布,幂指数的均值为1.38,有强阵发性和弱记忆性;用户回贴行为的波动性存在差异,利用熵理论对用户回贴行为波动有序性的量化表明,有序性越高,回贴行为的间间隔越不均匀,阵发现象越显著。在群体层面,大多主贴下群体用户回帖行为时间间隔分布服从幂律分布,幂指数的均值为1.36,有强阵发性和弱记忆性;以热门话题来表征群体回帖行为涌现,可以发现群体回帖行为涌现(热门话题爆发)的时间间隔分布服从幂律分布,有强阵发性和强记忆性。

【Purpose/significance】This paper explores the unique laws of information communication behavior in online community from the perspective of human dynamics,which will provide a theoretical basis for behavior modeling and service optimizing in online community.【Method/process】By applying and improving the classic analysis indexes of human dynamics,taking 79501 posts and related 979718 replies in Tianya community as the source data,this paper made a systematic quantitative analysis of reply behavior from the perspective of time interval,burstiness,memory,fluctuation and orderness,mainly includes the analysis of individual users’reply behavior,the analysis of group users’reply behavior under a single post and the emergence of group reply behavior(outbreak of hot topics)under a single post.【Result/conclusion】The results show that,at the individual level,the time interval of most users’reply behavior followed power-law distribution,and the mean value of the power exponent is 1.38,with strong burstiness and weak memory.There were individual differences in the fluctuation of users’reply behavior.The entropy theory was used to quantify the orderness of the user’s reply behavior fluctuation,it indicated that the higher the orderness,the more uneven the time interval,and the more obvious the intermittency.At the group level,the time interval of most posts which were replied followed power-law distribution,and the mean value of the power exponent is 1.36,with strong burstiness and weak memory.It can also be found that the time interval distribution of the emergence of group response behavior(outbreak of hot topics)followed power-law distribution,with strong burstiness and strong memory.