Journal of Frontiers of Computer Science and Technology ›› 2017, Vol. 11 ›› Issue (7): 1068-1079.DOI: 10.3778/j.issn.1673-9418.1606038

Previous Articles     Next Articles

User Relationships Prediction Algorithm with Interest Similarity Measurement

HUANG Hongcheng1,2+, LU Weijin1, HU Min1, WEI Qing1   

  1. 1. School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
    2. College of Computer Science, Chongqing University, Chongqing 400044, China
  • Online:2017-07-01 Published:2017-07-07

用户兴趣相似性度量的关系预测算法

黄宏程1,2+,陆卫金1,胡  敏1,魏  青1   

  1. 1. 重庆邮电大学 通信与信息工程学院,重庆 400065
    2. 重庆大学 计算机学院,重庆 400044

Abstract:  For the problem that current researchers only pay their attention to the changeability and ignore the durability when they study user interest in microblog, this paper proposes a user relationship prediction algorithm based on interest similarity. In this algorithm, interests are divided into two types: long-term interest characterized by labels and short-term interest characterized by texts, and a frequency statistics and multi-order quantization method is used to measure and update the degree of user interest according to the features of the interest. The similarity between user interests is computed by the method of cosine similarity which is used to predict user relationship. Results show that the proposed algorithm can accurately describe user??s interest, and improve the precision of user relationship prediction.

Key words: user interest, similarity, multistage quantization, relationship prediction

摘要: 针对目前研究微博用户兴趣变化时,只考虑用户兴趣的易变性而忽略了用户兴趣持久性的问题,提出了基于用户兴趣相似性的用户关系预测算法。将用户兴趣分为短期兴趣和长期兴趣,用户的文本信息表征为短期兴趣,用户的标签表征为长期兴趣。根据长短期兴趣的特征,采用频率统计和多阶量化的方法度量用户兴趣度并更新用户兴趣状态。最后通过余弦相似性指标计算用户间的兴趣相似度来预测用户关系。实验结果表明,该算法能够准确描述用户兴趣,提高用户关系预测的准确性。

关键词: 用户兴趣, 相似性, 多阶量化, 关系预测