Author Login Editor-in-Chief Peer Review Editor Work Office Work

Computer Engineering

Previous Articles     Next Articles

A Microblog Recommendation Method Based on Label Correlation Relationship

MA Huifang,JIA Meihuizi,LI Xiaohong,LU Xiaoyong   

  1. (College of Computer Science and Engineering,Northwest Normal University,Lanzhou 730070,China)
  • Received:2015-03-19 Online:2016-04-15 Published:2016-04-15

一种基于标签关联关系的微博推荐方法

马慧芳,贾美惠子,李晓红,鲁小勇   

  1. (西北师范大学计算机科学与工程学院,兰州 730070)
  • 作者简介:马慧芳(1981-),女,副教授、博士,主研方向为人工智能、数据挖掘、机器学习;贾美惠子,硕士研究生;李晓红,讲师;鲁小勇,讲师、博士研究生。
  • 基金资助:
    国家自然科学基金资助项目(61163039,61363058);甘肃省青年科技基金资助项目(145RJYA259);甘肃省自然科学研究基金资助项目(145RJZA232);西北师范大学2013年度青年教师科研能力提升计划基金资助项目(NWNU-LKQN-12-23);中国科学院计算技术研究所智能信息处理重点实验室开放基金资助项目(IIP2014-4)。

Abstract: A microblog recommendation method based on label correlation relationship is presented via analyzing mircoblog features and the deficiencies of existing microblog recommendation finding algorithms.Label retrieval strategy is adopted to add label for unlabeled users and users with small number of labels,and the user-label matrix is built and user-label weights are obtained.The label information is used to represent user interests.In order to solve the problem of high-dimension sparsity of the matrix,the correlation relationship between the labels is investigated to update the user-label matrix so as to get the final user interests.Experimental results show that compared with unlabeled microblog recommendation algorithm,the presented recommendation method is more effective in microblog information recommendation.

Key words: microblog recommendation, label retrieval, user-label matrix, user-label weight, label correlation relationship

摘要: 通过分析微博特点及现有微博推荐发现算法的缺陷,提出一种新的微博推荐方法。采用标签检索策略对未加标签和标签较少的用户进行加标,构建用户-标签矩阵,得到用户-标签权重并利用标签信息表征用户兴趣。为解决该矩阵中高维稀疏的问题,通过挖掘标签间的关联关系,继而更新用户-标签矩阵,获得最终的用户兴趣并进行相关推荐。实验结果表明,与忽略标签间关系的微博推荐方法相比,该推荐方法能够更有效地进行微博推荐。

关键词: 微博推荐, 标签检索, 用户-标签矩阵, 用户-标签权重, 标签关联关系

CLC Number: