在线社会网络中多话题竞态传播分析与建模  

Analysis and Modeling for Competitive Diffusion of Multiple Topics in Online Social Networks

作　　者：周亚东[1] 刘丽丽[1] 张贝贝[2] 雷蕾[3] 

机构地区：[1]西安交通大学智能网络与网络安全教育部重点实验室,西安710049 [2]西安理工大学计算机科学与工程学院,西安710048 [3]西安理工大学水利水电学院,西安710048

出　　处：《西安交通大学学报》2017年第2期1-5,39,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(61572397;61502383;61571360);陕西省自然科学基础研究计划资助项目(2015JM6298)

摘　　要：针对在线社会网络中多个话题在传播过程中呈现出的竞争状态,进行了竞态传播过程的测量与分析,并建立了多话题竞态传播模型。基于多个话题数据进行了话题参与用户行为分析,发现较少用户会持续关注同一个热点话题,并且会有一定数量用户在多个同类话题间转移关注,从而使得多个同类话题在并行传播时对吸引用户参与呈现出竞争态势。在分析结果的基础上,建立了考虑话题之间相互影响力以及话题吸引度的多话题竞态传播模型,该模型可有效描述多个同类热点话题在同时间段出现时各个话题之间的相互影响情况,以及各个话题在传播过程中人群参与规模的变化情况。在与实际数据的对比实验中,模型仿真结果的平均峰值出现时间的误差为0.2d,平均传播周期的误差为2.4d,话题间用户平均转出比例的误差为1.2%,并且能复现参与人数的单峰性、长尾特性等话题传播的动态特性。上述实验结果表明,该模型可有效描述在线社会网络中的多话题竞态传播动态过程。Competitive diffusion processes are analyzed, and a model for the competitive diffusion of multiple topics is proposed to understand the competitive diffusion of multiple topics in online social networks. Analyzing real network data, it is found that just a few users continue to focus on the same topic, and many users transfer their attentions among multiple similar topics. In this case, similar topics attract users to participate in a results, a diffusion model describing the dynamic competitive situation is proposed. Experiment competitive situation. Based on the analysis diffusion process of multiple topics in the validates that the model reproduces the unimodality and long tail characteristics of dynamic changes of users who participate in multiple topics, and achieves a good performance. A comparison with actual data show that the error of average peak time of the proposed model is 0.2 day, the error of average diffusion period is 2.4 days, and the error of average transferring proportion among topics is 1.2 %. These results show that the model can effectively describe the competitive diffusion of multiple topics in online social networks.

关 键 词：在线社会网络 热点话题 传播模型 用户行为分析 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]