作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 人工智能及识别技术 • 上一篇    下一篇

基于标签主题的协同过滤推荐算法研究

文俊浩 1,2,袁培雷 1,2,曾骏 1,2,王喜宾 1,2,周魏 1,2   

  1. (1.信息物理社会可信服务计算教育部重点实验室,重庆 400030; 2.重庆大学 软件学院,重庆 401331)
  • 收稿日期:2016-01-18 出版日期:2017-01-15 发布日期:2017-01-13
  • 作者简介:文俊浩(1969—),男,教授、博士,主研方向为数据挖掘、服务计算;袁培雷,硕士;曾骏,讲师、博士;王喜宾、周魏,博士。
  • 基金资助:
    国家自然科学基金(61379158,61502062);重庆市科技计划项目(cstc2014jcyjA40054)。

Research on Collaborative Filtering Recommendation Algorithm Based on Topic of Tags

WEN Junhao  1,2,YUAN Peilei  1,2,ZENG Jun  1,2,WANG Xibin  1,2,ZHOU Wei  1,2   

  1. (1.Key Laboratory of Dependable Service Computing in Cyber Physical Society,Ministry of Education,Chongqing 400030,China;2.School of Software Engineering,Chongqing University,Chongqing 401331,China)
  • Received:2016-01-18 Online:2017-01-15 Published:2017-01-13

摘要: 传统基于标签的推荐算法仅考虑用户的评分信息,导致推荐准确度不高。为解决该问题,提出一种改进的协同过滤推荐算法。对用户-标签矩阵、资源-标签矩阵进行潜在Dirichlet分布建模,发掘推荐系统中的潜在语义主题,从语义层面计算用户对各资源的偏好概率,将计算出的偏好概率与协同过滤算法计算出的资源相似度相结合,预测用户偏好值,实现个性化推荐。在Movielens数据集上的实验结果表明,与传统基于标签的推荐算法相比,该算法能消除标签中存在的同义词、多义词等语义模糊问题,同时提高推荐准确度。

关键词: 标签主题, 协同过滤, 潜在Dirichlet分布模型, 个性化推荐, 相似度

Abstract: Aiming at the problem that traditional tag-based recommendation algorithm only considering the user’s rating information leads to the low accuracy of recommendation,this paper puts forward an improved collaborative filtering recommendation algorithm.It builds the Latent Dirichlet Allocation(LDA) model for user-tag matrixes and resource-tag matrixes and explores the latent semantic topics in the recommendation system.The probability of users choosing resources from the semantic level is computed.Then the probability is combined with resource similarity calculated through collaborative filtering algorithm to predict user preference values to realize personalized recommendation.Experimental results on the Movielens dataset show that the proposed algorithm can eliminate the semantic ambiguities such as synonyms and polysemes,and improve the accuracy of recommendation,compared with the traditional tag-based recommendation algorithm.

Key words: topic of tag, collaborative filtering, Latent Dirichlet Allocation(LDA) model, personalized recommendation;similarity

中图分类号: