基于自激点过程的网络热点话题传播模型  

An Efficient Topic Propagation Model Based on Self-Exciting Point Process

作　　者：韩忠明[1] 张梦[1] 谭旭升 段大高[1] 司慧琳[1] 

机构地区：[1]北京工商大学计算机与信息工程学院,北京100048

出　　处：《计算机学报》2016年第4期704-716,共13页Chinese Journal of Computers

基　　金：国家自然科学基金(61170112);教育部人文社会科学研究青年基金(13YJC860006);中央财政支持地方高校发展专项资金人才培养和创新团队建设项目(19005323132)资助

摘　　要：建模互联网中热点话题的传播过程具有重要的意义和价值,该文以网络热点话题为研究对象,基于自激霍克斯过程提出一个话题传播模型(Self-Exciting Point Process Model,SEPPM).SEPPM利用用户参与话题的自激效应,将话题传播过程建模为一个随机点过程.同时,SEPPM也考虑了话题传播的外部因素,综合形成话题传播模型.为了验证该模型的有效性,该文从仿真和实证两个角度分别进行了大量的实验比较,提出话题仿真算法,仿真结果说明SEPPM可以生成多种符合热点话题传播特征的模式.实际数据上的结果说明SEPPM不仅能够很好地拟合真实话题的传播过程,还能够有效地预测话题传播趋势.Modeling propagation processes of hot topics on Internet has significant meaning and value.This paper focuses on modeling hot topics on Internet and proposes a topic propagation model（Self-Exciting Point Process Model,SEPPM）based on self-exciting Hawkes process.SEPPM models the propagation process of one topic as a random point process by using self-exciting effect of user participation.At the same time,SEPPM also takes external factors for propagation into account,thus puts forward a formal topic propagation model.To evaluate effectiveness of SEPPM,comprehensive simulation and empirical experiment are conducted.A simulation algorithm for SEPPM is proposed and the simulated results show that SEPPM can generate a variety of patterns of hot topics with different propagation characteristics.The experimental results on real datasets show that SEPPM can not only splendidly fit real topic propagation process,but also can effectively forecast spreading trend.

关 键 词：社会媒体 热点话题 传播模型 霍克斯过程 自激点过程 社交网络 数据挖掘 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]