基于分层的微博话题传播模型研究

Mircroblog Diffusion Model Based on Hierarchical

通过对社交网络新浪微博的数据的统计分析,得知微博数据具有高度的聚集性,即一个流行微博的只被转发一次的转发数占总转发数量的50%以上.因此,提出了对信息级联分层的STIC模型,该模型的第一层级联和第二层级联分别使用SVM分类算法和基于主题的信息级联模型对话题传播进行预测.实验结果表明,STIC模型的预测结果优于基于主题的信息级联模型.

In the report, the statistical analysis of social network data from Sina Microblog was perforrp, ed. The a- nalysis data suggested that the aggregation of data from microblog was high, and the number of on time forward- ing was above 50% of the total number of forwarding. So, a hierarchical model, namely STIC, was proposed, and in which the information cascade was divided into two layers. The first layer was applied to describe the situ- ations where messages are forwarded only one time. The second layer was applied to describe the other forward-ing patterns. In the first layer, SVM was used to predict the information diffusion; In the second layer, Inde- pendent Cascade Model was used to predict the information diffusion. The results showed that the performance of STIC was superior to that of TIC in terms of accuracy of information diffusion.