基于受限非负张量分解的用户社会影响力分析

User social influence analysis based on constrained nonnegative tensor factorization

针对传统社会影响力分析方法未能充分考虑观点和话题信息等问题,提出了一种基于受限非负张量分解的用户社会影响力分析方法。首先把社交媒介用户相互评论关系自然地表示成三阶张量,然后通过拉普拉斯话题约束矩阵控制张量分解过程,最后根据分解得到的潜在因子度量用户观点社会影响力。该方法的优点是能有效地从受限张量分解结果中检索出给定话题下用户的社会影响力,同时保持其社会影响力的极性分布。实验结果表明,该方法的性能优于OOLAM和Twitter Rank等基准算法。

Existing models for measuring user social influence fail to integrate both opinion and topic information. Therefore, a new constrained nonnegative tensor factorization method combining user＇s opinion and the topical relevance was proposed. The method represented user＇s comment relations as 3-order tensor, factorized the comments tensor constrained by Laplacian topical matrix, and then measures user influence according to the latent factors resulting from the tensor factorization. Thus, the new method not only was capable to effectively calculate the strength of user social influence on given topic, but also kept the polarity allocation of social influence. The experimental result shows that the performance of the proposed method is better than that of the baseline methods such as OOLAM, Twitter Rank, etc.