融合多层图与分类信息的双意图会话推荐  

Dual intent session-based recommendation integrating multi-layer graphs and classification information

在线阅读下载全文

作  者:刘超[1] 王中迪 余岩化 朱军 Liu Chao;Wang Zhongdi;Yu Yanhua;Zhu Jun(School of Computer Science&Engineering,Chongqing University of Technology,Chongqing 400054,China)

机构地区:[1]重庆理工大学计算机科学与工程学院,重庆400054

出  处:《计算机应用研究》2025年第4期1058-1064,共7页Application Research of Computers

基  金:2021年重庆市社会科学规划一般项目(2021NDYB101);2023年重庆市教育委员会人文社会科学研究青年项目(23SKGH264);2024年重庆理工大学高质量发展行动计划资助项目(gzlcx20243200)。

摘  要:针对现有会话推荐系统存在的会话间信息挖掘不够充分、会话间聚合信息冗余和辅助信息未与会话特征相结合的问题,提出融合多层图与分类信息的双意图会话推荐模型(SRIMC)。首先,根据会话序列,构建局部会话图、会话关系图和全局项目图,通过图神经网络(GNN)学习得到局部会话特征、会话关系特征和全局项目会话特征,并将上述特征结合获得α意图;其次,基于替换先验分布为β分布的贝叶斯分布整合分类信息与会话长度信息,获得β意图;最后,将α和β意图融合进行预测。在五个公开数据集上的实验结果表明,SRIMC的P@20提升了1.23%~51.78%,MRR@20提升了2.87%~80.87%,证明了模型利用多层会话信息与分类信息捕获用户意图的有效性。Aiming to address the issues in existing session-based recommendation systems,such as insufficient mining of information between sessions,redundant aggregation of information between sessions,and the lack of integration of auxiliary information with session features,this paper proposed a dual intent session-based recommendation model integrating multi-layer graph and classification information model(SRIMC).Firstly,based on session sequences,the paper constructed a local session graph,a session relationship graph,and a global project graph.By using GNN,the method learned the local session features,session relationship features,and global items session features,and combined these features to obtain the α-intent.Next,by integrating classification information and session length information using a Bayesian distribution that replaced the prior with a β-distribution,the method obtained the β-intent.Finally,α-intent andβ-intent were combined for prediction.Experimental results on five public datasets show that the SRIMC model improves P@20 by 1.23% to 51.78% and MRR@20 by 2.87% to 80.87%,demonstrating the model’s effectiveness in capturing user intent by leveraging multi-layer session information and classification information.

关 键 词:会话推荐 多层信息 图神经网络 分类信息 双意图 

分 类 号:TP391[自动化与计算机技术—计算机应用技术]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象