摘要
利用推荐系统进行群组推荐时,群组成员之间的交互关系对推荐结果有很大影响,但传统的群组推荐算法较少考虑用户信任度的重要性,致使社交关系信息不能得到充分利用。在群组融合时考虑群组内用户间的交互关系,提出一种基于用户信任度和概率矩阵的群组推荐算法。在获取用户信任度数据后,使用概率矩阵分解(PMF)算法补全信任度矩阵并进行归一化处理,得到相似度矩阵,同时在后验概率计算过程中加入用户间的信任度因素,通过极大化后验概率获得预测评分。在此基础上,对群组中用户的权重进行归一化处理,使用基于用户交互关系的权重策略融合群组成员偏好,得到最终的推荐结果。在Epinions和FilmTrust数据集上的实验结果表明,该算法可使融合结果更具群组特性,同时提高推荐结果的可靠性和可解释性,且均方根误差和命中率均优于PMF、NeuMF、RippleNet等对比算法。
When using the recommendation system for group recommendation,interactions between group members have a great impact on recommendation results,but traditional group recommendation algorithms consider little the importance of user trust,and fail to make full use of social relationship information.Based on probability matrix and user trust,this paper proposes a group recommendation algorithm that considers interactions between users in a group.After obtaining data of user trust,the Probability Matrix Factorization(PMF)algorithm is used to complete the trust matrix,which is then normalized to obtain the similarity matrix. The factor of trust between users is added in the posterior probability calculation process to obtain the prediction score by maximizing the posterior probability. On this basis,the weights of users in the group are normalized,and a weight strategy based on user interactions is used to fuse the preferences of group members to obtain the recommendation result. The experimental results on Epinions dataset and FilmTrust dataset show that the proposed algorithm can make the fusion results to have more group characteristics,and improve the reliability as well as interpretability of the recommendation results. The algorithm displays a lower root mean square error and higher hit rate than PMF,NeuMF,RippleNet and other comparison algorithms.
作者
宋玉龙
马文明
刘彤彤
SONG Yulong;MA Wenming;LIU Tongtong(School of Computer and Control Engineering,Yantai University,Yantai,Shandong 264005,China)
出处
《计算机工程》
CAS
CSCD
北大核心
2022年第1期105-111,共7页
Computer Engineering
基金
国家自然科学基金(61602399)。
关键词
群组推荐
偏好融合
概率矩阵分解
用户信任度
协同过滤
group recommendation
preference fusion
Probability Matrix Factorization(PMF)
user trust
Collaborative Filtering(CF)
