基于用户潜在信任与被信任关系的推荐系统  

Recommender Systems Based on User Latent Trust and Trusted Relationship

作　　者：于佳玄 李明 丁德锐 YU Jiaxuan;LI Ming;DING Derui(College of Science,University of Shanghai for Science and Technology,Shanghai 200093;School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093)

机构地区：[1]上海理工大学理学院,上海200093 [2]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机与数字工程》2025年第3期652-659,共8页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61973219)资助。

摘　　要：由于用户与条目数的急剧增加,以及为充分考虑用户之间的潜在关联信息,现有的推荐系统模型相对保守。因此,论文提出了一种考虑用户之间潜在信任与被信任关系的推荐系统模型。该模型具有以下特点:1)引入图拉普拉斯正则化确保了潜在信任空间中的用户结构信息,进而得到了用户之间的潜在信任与被信任关系矩阵;2)将得到关系信息迁移到评分矩阵的分解过程中,并引入相似用户局部学习社交正则化,实现了相似用户对评分条目喜好的一致性;3)利用单因素方法设计了变量更新原则,降低了由于知识迁移等所导致的计算复杂度。最后,在六个真实的数据集上的实验结果表明,论文所提出的模型在保证计算效率的同时能够获得较高的预测精度。The existed recommender system models are relatively conservative due to the dramatic increase in the number of users and entries,and the potential correlation information between users without full consideration.As such,this paper proposes a novel model of recommender systems based on that utilizes the latent trust and trusted relationship between users.Specifically,the model has the following characteristics.The matrix of potential trust and trusted relationship between users can be obtained by introducing the graph Laplacian regularization to ensure the user structure information in the latent trust space.The consistency of similar users'preferences for scoring items is ensured via knowledge transfer of the obtained relationship matrix to the decomposition process of the scoring matrix as well as the utilization of the locally learned social regulation of similar users,and the computational complexity caused by knowledge transfer is effectively reduced by resorting to a new single-factor-based variable update principle.Finally,the experimental results on six real datasets show that the model proposed in this paper can achieve high prediction accuracy while ensuring computational efficiency.

关 键 词：推荐系统 潜在信任关系 潜在被信任关系 图拉普拉斯正则化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]