社交电商中融合信任和声誉的图神经网络推荐研究

社交电商可依据用户间的社交关系为用户提供感兴趣的商品或服务。现有研究多基于社会信任或社会声誉进行推荐，却忽略了信任与声誉间的相互作用，导致推荐效果欠理想。针对以上问题，本文提出了一种融合信任(Trust)和社会声誉(Social Reputation)的图神经网络推荐算法(TSR-GM)，采用社会声誉来深度刻画用户关系在推荐系统中的作用，利用社交网络中用户被信任程度对用户声誉进行排名，以图神经网络量化整合用户信任与声誉，并将结合后的新矩阵不断校正以获取更准确的用户信任，以此对矩阵分解后得到的新评分模型更新，最终得到更准确度量的预测评分矩阵。运用Epinions数据集开展的相关实验表明：与同类方法比，TSR-GM算法对提高推荐精度有较好效果。

A Graph Neural Network Recommendation Study Combining Trust and Reputation in Social E-commerce

Amid rapid development of social network, growing number of users leads to excessive information, challenging the beneficiaries in theory to filter the valuable bits. One question raised along this, which requires users to identify trustworthy content, presents researchers with the idea of studying the role of which trust relations among users play in social e-commerce. Some exploring in this field sees a connection between item ratings and trust relations of users, however, there often lacks integrity in reliable trust relation data, and even those that doesn’t show great disparity. Hence, a method to identify content and information that users trusted more swiftly and precisely is still in great demand. Also, researchers show increasing interest in the position that user reputation played in effecting user trust and how to merge those two to offer more satisfying recommendation.