北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2013, Vol. 36 ›› Issue (1): 59-62.doi: 10.13190/jbupt.201301.59.156

• 论文 • 上一篇    下一篇

综合时间序列分析和新颖性的信息扩散预测

蔡飞,陈洪辉,舒振   

  1. 信息系统与管理学院 信息系统工程重点实验室, 长沙 410073
  • 收稿日期:2012-04-22 修回日期:2012-10-18 出版日期:2013-02-28 发布日期:2013-01-19
  • 通讯作者: 蔡飞 E-mail:caifei@nudt.edu.cn
  • 作者简介:蔡飞(1984-),男,博士生,Email:caifei@nudt.edu.cn 陈洪辉(1969-),男,教授,博士生导师
  • 基金资助:

    国家自然科学基金项目(61070216)

Information Diffusion Prediction Based on Time-Series Analysis and Information Novelty

CAI Fei, CHEN Hong-hui, SHU Zhen   

  1. Science and Technology on Information Systems Engineering Laboratory, College of Information System and Management, Changsha 410073, China
  • Received:2012-04-22 Revised:2012-10-18 Online:2013-02-28 Published:2013-01-19
  • Contact: Fei CAI E-mail:caifei@nudt.edu.cn

摘要:

现有信息扩散预测普遍依赖于社会网络构建,从而引发网络链路估计准确率低,信息扩散预测精度差的问题,为此提出了一种综合时间序列分析和信息新颖性的信息扩散预测方法.通过分析信息在网络节点上扩散随时间的变化特性,对网络节点的全局影响力进行估计,并考虑信息产生至节点受影响的时间差来衡量信息新颖性,进而平移调整节点影响力大小,最终实现信息扩散范围的预测.向斯坦福大学所提供测试数据的实验结果表明,新方法准确稳定地预测了信息扩散范围的实时变化.

关键词: 信息扩散预测, 时间序列分析, 信息新颖性, 影响传播

Abstract:

To solve the problem of low precision on information propagation prediction caused by frustrated link prediction in social networks, a new approach, based on time-series analysis, combined with information novelty, is proposed with no requiring the knowledge of social network.When the node is infected, the global influence function of a node is estimated. The overall propagation volume of information is predicted. Simulations on dataset about news articles and blogs show that the proposal can accurately and reliably predict the temporal variations of information diffusion.

Key words: information diffusion prediction, time-series analysis, information novelty, influence propagation

中图分类号: