SGRec:一种基于双层信息交互的会话推荐算法  

SGRec:a Session-based Recommendation Algorithm on Two-layer Information Interaction

作　　者：王誉熹 彭敦陆[1] WANG Yuxi;PENG Dunlu(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《小型微型计算机系统》2024年第6期1392-1397,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61772342)资助。

摘　　要：会话推荐是根据匿名用户的交互序列去推荐该用户下一个最有可能交互的项目.在现有的会话推荐模型中,大多数模型只学习了图的单层信息,这种学习方式会导致对交互序列的信息提取不完整.本文提出了一种结合项目的图级信息与序列级信息的推荐算法.图级信息是将用户的交互序列映射为一个高维空间超图,通过超图神经网络去学习图中每个节点的信息;会话中项目的序列级信息采用深度序列提取器和注意力网络去获取,最终将两组信息融合并通过自注意力网络进行下一项推荐.通过这种方法可以获得会话序列中每个项目更完整的信息.本文在真实数据集Diginetica,Tmall,Nowplaying上设置对比实验验证了算法的有效性,该算法在MRR@N和P@N上有明显提升,有效地证明了本文算法的推荐性能.Session-based recommendation aims to recommend the item which is most likely to interact with next based on the interaction sequence of anonymous users.In the existing session-based recommendation models,most of them only learn the single layer information of the graph,which will not complete the information extraction of the interaction sequence.This paper presents a recommendation algorithm that combines graph-level information and sequence-level information.Graph-level information is to map the user interaction sequence into a high-dimensional hypergraph,and learn the information of each node in the graph through the hypergraph neural network.The sequence-level information of the item in the session is obtained by the deep sequence extractor and the attention network.Finally,the two groups of information are fused and the next recommendation is made through the self-attention network.With this approach,it can get more complete information about each item in the session sequence.In this paper,a comparative test is set on the real datasets Diginetica,Tmall,Nowplaying to verify the effectiveness of the algorithm.The algorithm has significantly improved in MRR@N and P@N,effectively proving the recommendation performance of the algorithm in this paper.

关 键 词：会话推荐 超图神经网络 注意力网络 循环神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]