一种基于多模式影响力的微博信息扩散预测方法

摘要

Microblogs have become an essential way to improve the communications among people.If the outbreak of a mes-sage could be predicted,it would bring in valuable opportunities or devastating effects.Also,it would be important if the people could predict whether a message would become viral,and which users would participate in spreading this message.This paper defined three types of influences,such as interest-oriented influence,social-oriented influence,and epidemic-oriented influ-ence.Then,it proposed a diffusion-targeted influence model to differentiate and quantify various types of influence.Further-more,it modeled the problem of diffusion prediction by factorizing a user’s intention to transmit a microblog into these influ-ences.It could predict the future diffusion state of any new microblog by learned prediction model.A real-world microblogs dataset was conducted to evaluate this method.The results demonstrate the superiority of the model.%微博服务已经成为加强人们互相沟通的重要媒体。微博的信息扩散能够带来巨大的商业价值，同时也为谣言的传播提供了良好滋润环境。如果能够预测某个信息是否会被爆发式地传播，以及哪些用户会加入到信息的传播中将是非常有意义的工作。定义了三种模式的影响力：面向兴趣的影响力、面向社交关系的影响力和面向传播能力的影响力，提出了一种基于扩散的影响力模型来区分和度量不同类型的影响力；进一步地，利用微博用户传播博文给影响力用户的意图来对影响力传播预测建模，预测模型能够预测新博文将来的扩散状态。利用真实的微博数据来评测预测模型，实验结果表明提出的方法准确性优于传统的预测模型。