Journal of Computer Applications ›› 2019, Vol. 39 ›› Issue (3): 634-638.DOI: 10.11772/j.issn.1001-9081.2018071521

Previous Articles     Next Articles

Collaborative filtering recommendation algorithm based on tag weight

LEI Man1,2, GONG Qin1,2, WANG Jichao1,2, WANG Baoqun1,2   

  1. 1. School of Communication and Information Engineering, Chongqing University of Posts and Communications, Chongqing 400065, China;
    2. Chongqing Key Lab of Mobile Communications Technology(Chongqing University of Posts and Communications) Chongqing 400065, China
  • Received:2018-07-23 Revised:2018-09-17 Online:2019-03-10 Published:2019-03-11
  • Contact: 雷曼
  • Supported by:
    This work is partially supported by the Program for Changjiang Scholars and Innovative Research Team in University (IRT1299).


雷曼1,2, 龚琴1,2, 王纪超1,2, 王保群1,2   

  1. 1. 重庆邮电大学 通信与信息工程学院, 重庆 400065;
    2. 移动通信技术重庆市重点实验室(重庆邮电大学), 重庆 400065
  • 作者简介:雷曼(1992-),女,重庆人,硕士研究生,主要研究方向:推荐算法、数据分析;龚琴(1993-),女,四川成都人,硕士研究生,主要研究方向:深度学习、自然语言处理;王纪超(1993-),男,河南省郑州人,硕士研究生,主要研究方向:机器学习、数据挖掘;王保群(1993-),男,山东省聊城人,硕士研究生,主要研究方向:机器学习、数据挖掘。
  • 基金资助:

Abstract: Aiming at the problem that the recommendation accuracy is not good enough due to the similarity calculation in traditional collaborative filtering recommendation algorithm, a collaborative filtering recommendation algorithm based on the similarity measurement method of tag weight was proposed. Firstly, the calculation of tag weights in existing algorithm was improved to construct a user-tag weight matrix and an item-tag weight matrix. Secondly, as the recommendation system is based on the user-centered recommendation, the most accurate evaluation and demand of the users were obtained by constructing a user-item association matrix. Finally, according to the user-item bipartite graph, the similarity between users based on the label weight was calculated by the material diffusion algorithm, and the recommendation lists were generated for the target users. The experimental results show that compared with UITGCF (a hybrid Collaborative Filtering recommendation algorithm by combining the diffusion on User-Item-Tag Graph and users' personal interest model), when the sparsity environment is 0.1, the recall, accuracy, F1 score of the proposed algorithm were respectively increased by 14.69%, 9.44% and 17.23%. When the recommendation item number is 10, the three indicators respectively were increased by 17.99%, 8.98%, and 16.27%. The results show that the collaborative filtering recommendation algorithm based on tag weight effectively improves the recommendation results.

Key words: user-tag weight, item-tag weight, recommendation system, collaborative filtering, material diffusion

摘要: 针对传统协同过滤推荐算法中由于相似度计算导致推荐精度不足的问题,提出一种基于标签权重相似度量方法的协同过滤推荐算法。首先,通过改进当前算法中标签权重的计算,并构成用户-标签权重矩阵和物品-标签权重矩阵;其次,考虑到推荐系统是以用户为中心进行推荐,继而通过构建用户-物品关联矩阵来获取用户对物品最准确的评价和需求;最后,根据用户-物品的二部图,利用物质扩散算法计算基于标签权重的用户间相似度,并为目标用户生成推荐列表。实验结果表明,与一种基于"用户-项目-用户兴趣标签图"的协同好友推荐算法(UITGCF)相比,在稀疏度环境为0.1时该算法的召回率、准确率和F1值分别提高了14.69%、9.44%、17.23%。当推荐项目数量为10时,三个指标分别提高了17.99%、8.98%、16.27%。结果表明基于标签权重的协同过滤推荐算法可有效提高推荐结果。

关键词: 用户-标签权重, 物品-标签权重, 推荐系统, 协同过滤, 物质扩散

CLC Number: