张量分解在用户影响力度量中的应用
The application of tensor factorization on user influence measure

提出一种基于张量分解的有影响力用户识别算法. 该算法首先构建基于查询主题的用户交互关系张量，接着利用张量分解算法对用户行为进行预测，最后融合各种交互关系和用户的主题信息给出用户影响力的综合评判. 实验结果表明，与非负矩阵分解相比，张量分解的挖掘精度提升了约10%，而与PageRank相比，张量分解的挖掘精度提升了约20%.
This paper proposes a method for identifying influential users based on tensor factorization (TF). This method fitst constructs interaction tensor based on the topic. Then，we extract the user behavior by factorizing the tensor. Finally，the users’ influences are evaluated by comprehensively considering the topic and interaction information between users. The experiments shows that，compared with NMF and PageRank method，the accuracy of TF method can increase by 10% and 20%，respectively