Loading…

Global and session item graph neural network for session-based recommendation

Session-based recommendation algorithm is a research hotspot with economic significance and research value. Most of the algorithms are based on how to represent users and items better. Deep learning has made many achievements in the recommendation area due to its strong representation ability. Never...

Full description

Saved in:
Bibliographic Details
Published in:Applied intelligence (Dordrecht, Netherlands) Netherlands), 2023-05, Vol.53 (10), p.11737-11749
Main Authors: Sheng, Jinfang, Zhu, Jiafu, Wang, Bin, Long, Zhendan
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Session-based recommendation algorithm is a research hotspot with economic significance and research value. Most of the algorithms are based on how to represent users and items better. Deep learning has made many achievements in the recommendation area due to its strong representation ability. Nevertheless, in recent years, the excellent performance of the graph neural network in network representation provides many inspirations. Numerous recommendation algorithms based on graph neural networks only consider building a graph for each session to handle recommendation tasks. The relationship between items in the entire data set is ignored. Therefore, we propose a recommendation algorithm called GS-GNN for integrating items’ global features and local features by graph neural network. By modeling the entire data as a global graph, we use the graph attention network to learn global representations of the items. We model each session as a session graph and use a gated graph neural network to learn local representations of the items. Sessions’ representations are obtained through the fusion of items. The task is to recommend top-k items for each session by items’ and sessions’ representations. We did a comparative experiment and performance analysis experiment. The comparative experiments prove the effectiveness of GS-GNN, and we also conduct a detailed analysis of the model through experiments.
ISSN:0924-669X
1573-7497
DOI:10.1007/s10489-022-04034-w