摘要
随着互联网信息的发展,网络数据量大幅增长,极大提高了用户的有效信息筛选难度。推荐系统根据用户的历史行为和偏好信息而产生相应的推荐,协同过滤算法是推荐系统中的一种常用算法。传统的协同过滤算法仅使用相似度作为推荐依据时,仍然面临推荐精确率不高问题,本文在相似度基础上添加用户之间的信任度,对用户之间不对等的信任关系建模,再添加对热门项目的惩罚机制,从而弱化热门项目的推荐。通过对MovieLens数据集的实验结果进行验证可知,融合信任度的协同过滤算法的精确率、覆盖率和F1值均比传统的基于用户的协同过滤算法性能有所提高。
With the development of Internet information,the amount of network data increases greatly,and so does the difficulty of effective information screening for users.A recommendation system generates recommendations based on the historical behaviors and preferences of users.Collaboration filtering is a common algorithm in the recommendation system.When the traditional collaborative filtering algorithm only uses similarity as the recommendation basis,it still faces the problem of low recommendation accuracy.In this paper,the trust between users is added on the basis of similarity,the trust relationship between users is modeled,and the punishment mechanism for popular items is added,so as to weaken the recommendation of popular projects.Verified by experimental results on the MovieLens dataset,the TrustUserCF-Based collaborative filtering algorithm has improved the performance of precision,coverage and F1-score compared with the traditional user-based collaborative filtering algorithm.
作者
何进成
王浩
孙刚
刘其刚
HE Jincheng;WANG Hao;SUN Gang;LIU Qigang(School of Computer and Information Engineering,Fuyang Normal University,Fuyang 236037,China)
出处
《长春师范大学学报》
2024年第4期54-59,91,共7页
Journal of Changchun Normal University
基金
国家自然科学基金项目“融合实体关系模型的视觉问答关键技术研究”(61906044)
安徽省教育厅自然科学研究项目“基于半监督学习的不完全标记文本数据流的分类问题研究”(KJ2020ZD48)。
关键词
推荐系统
协同过滤
相似度
信任度
热门惩罚
recommendation system
collaborative filtering
similarity
trust
popular punishment
