• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 人工智能与数据挖掘 • 上一篇    下一篇

融合邻居选择策略和信任关系的兴趣点推荐

刘辉1,2.3,曾斌1,2,刘子恺4   

  1. (1.重庆邮电大学通信与信息工程学院,重庆 400065;2.重庆邮电大学通信新技术应用研究中心,重庆 400065;
    3.重庆信科设计有限公司,重庆  401121;4.重庆邮电大学计算机科学与技术学院,重庆 400065)
     
  • 收稿日期:2019-06-04 修回日期:2019-08-05 出版日期:2020-02-25 发布日期:2020-02-25

Point-of-interest recommendation based on
neighbor selection strategy and trust relationship
 

LIU Hui1,2,3,ZENG Bin1,2,LIU Zi-kai4   

  1. (1.School of Communication and Information Engineering,
    Chongqing University of Posts & Telecommunications,Chongqing 400065;
    2.Research Center of New Telecommunication Technology Applications,
    Chongqing University of Posts & Telecommunications,Chongqing 400065;
    3.Chongqing Information Technology Designing Co.,Ltd.,Chongqing 401121;
    4.School of Computer Science and Technology,Chongqing University of
    Posts & Telecommunications,Chongqing 400065,China)
  • Received:2019-06-04 Revised:2019-08-05 Online:2020-02-25 Published:2020-02-25

摘要:

兴趣点推荐是推荐系统的关键研究之一,传统的算法只利用用户签到信息进行推荐,且对于签到信息只单纯地考虑签到和没签到,而忽略了用户签到的频次和信任关系。为提高推荐精度,提出了一种融合用户相似性、地理位置和信任关系的混合推荐算法(UGT)。对于签到信息,采用签到频次来代替传统的二值签到,并对签到信息添加时间权重;对于基于用户的协同过滤,提出了一种邻居选择策略来提高预测精度;对于信任关系,首先分析用户的属性,然后给出社会地位的计算方法,重构信任度的计算方法。实验结果表明,该混合算法相比较传统的推荐算法而言,在准确率和召回率上有了显著的提升。
 

关键词: 位置社交网络, 兴趣点推荐, 协同过滤, 信任关系, 时间权重

Abstract:

Point-of-interest recommendation is one of the key researches in recommendation systems. Traditional algorithms only use user sign-in information for recommendation. They only consider whether users sign in or not, and ignore the frequency of user sign-in and trust relationship. In order to improve the recommendation accuracy, a hybrid recommendation algorithm (UGT) combining user similarity, geographic location and trust relationship is proposed. For sign-in information, sign-in frequency is used to replace the traditional binary sign-in, and the time weight is added to the sign-in information. For user-based collaborative filtering, a neighbor selection strategy is proposed to improve the prediction accuracy. For trust relationship, the user’s attributes are firstly analyzed, then a social status calculation method is given, and a trust degree calculation method is reconstructed. Experiments show that the hybrid algorithm significantly improves accuracy and recall rate in comparison to the traditional recommendation algorithm.
 

Key words: