基于会话的结合全局潜在信息的图神经网络推荐模型  

Global potential information combined graph neural networks forsession-based recommendation

作　　者：董立岩[1,2] 梁伟业 王越群[1] 李永丽 DONG Li-yan;LIANG Wei-ye;WANG Yue-qun;LI Yong-li(College of Computer Science and Technology,Jilin University,Changchun 130012,China;Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education,Jilin University,Changchun 130012;School of Computer Science and Technology,Northeast Normal University,Changchun 130117,China)

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]东北师范大学信息科学与技术学院,长春130117

出　　处：《吉林大学学报（工学版）》2023年第10期2964-2972,共9页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(61872164);吉林省科技发展计划项目(20190302032GX)。

摘　　要：针对当前各种基于会话推荐的模型存在对物品间全局关系的获取和使用不足的问题,提出了基于会话的结合全局潜在信息的图神经网络推荐模型。该模型根据全部会话序列构建会话图与全局图,并在全局图中引入了序列中各节点间的间距信息,以及序列节点的相邻节点彼此之间的贡献度,通过模型训练获取最后的会话表征预测下一个交互行为。实验结果表明:在结合图神经网络的推荐算法中充分挖掘全局潜在信息可以有效提高推荐算法的准确率,这一改进对提高基于会话的图神经网络模型的性能有一定指导意义。Aiming at the problem of insufficient acquisition and use of the global relationship between items in various current session based recommendation models,a session based graph neural network recommendation model combined with global potential information is proposed.The model constructs session graph and global graph according to all session sequences,introduces the spacing information between nodes in the sequence and the contribution of adjacent nodes of sequence nodes to each other,obtains the final session representation through model training,and predicts the next interaction behavior.The experimental results show that fully mining the global potential information in the recommendation algorithm combined with graph neural network can effectively improve the accuracy of the recommendation algorithm.This improvement has certain guiding significance for improving the performance of session based graph neural network model.

关 键 词：计算机软件 推荐系统 基于会话的推荐 图神经网络 全局信息 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]