Graph Isomorphism and Hybrid-Order Residual Gated Graph Neural Network for Session-Based Recommendation

Session-based recommendation aims to predict which item will be clicked next for the current session. Aiming at the shortcomings of existing session recommendation models based on graph neural network, this paper proposes a model named graph isomorphism and hybrid-order residual gated graph neural n...

Full description

Saved in:
Bibliographic Details
Main Author: WANG Yonggui, YU Qi
Format: Article
Language:zho
Published: Journal of Computer Engineering and Applications Beijing Co., Ltd., Science Press 2025-02-01
Series:Jisuanji kexue yu tansuo
Subjects:
Online Access:http://fcst.ceaj.org/fileup/1673-9418/PDF/2401029.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Session-based recommendation aims to predict which item will be clicked next for the current session. Aiming at the shortcomings of existing session recommendation models based on graph neural network, this paper proposes a model named graph isomorphism and hybrid-order residual gated graph neural network for session-based recommendation (GIHR-GNN). Firstly, graph isomorphic network is used to aggregate feature vectors of adjacent items, which can effectively fuse global and local information, solving the problem that graph neural network is good at capturing local connections between nodes and ignoring global information. And the user􀆳s long-term and short-term interests are aggregated by gated fusion to capture dynamic changes of user interests. Secondly, hybrid-order gated graph neural network is used to process the position embedding to capture the user intention reflected by the user􀆳s re-interaction after a long time, and the residual module is added to solve the degradation problem of deep network. Finally, contrastive learning is applied to comparing the user􀆳s long-term interest representations without denoising and after denoising, alleviating the problems of data sparsity and noise interference. Experiments are performed on Tmall and RetailRocket datasets. Compared with baseline model, GIHR-GNN increases at least 3.26% and 10.33% in P@20 and MRR@20 on Tmall, 0.55% and 2.57% in P@20 and MRR@20 on RetailRocket, which proves the effectiveness of GIHR-GNN.
ISSN:1673-9418