Accessibility navigation


Self-supervised enhancement method for multi-behavior session-based recommendation

Zhang, Z., Zhang, J. ORCID: https://orcid.org/0009-0007-7012-7540, Chen, J., Huang, Y. and Huang, X. (2024) Self-supervised enhancement method for multi-behavior session-based recommendation. IEEE Access, 12. pp. 175268-175277. ISSN 2169-3536

[thumbnail of Open Access]
Preview
Text (Open Access) - Published Version
· Available under License Creative Commons Attribution.
· Please see our End User Agreement before downloading.

980kB

It is advisable to refer to the publisher's version if you intend to cite from this work. See Guidance on citing.

To link to this item DOI: 10.1109/access.2024.3504496

Abstract/Summary

Session-Based Recommendation(SBR) aims to capture users’ short-term and dynamic preferences through anonymous sessions. Most existing SBR methods neglect the collaborative information between multiple behaviors in a session when modeling user preferences, and they often struggle to capture the complex correlations of contextual information across sessions, which can lead to poor recommendation performance. To address these issues, this paper proposes a Self-Supervised enhancement method for Multi-Behavior Session-based Recommendation(SSMB-SR). SSMB-SR represents the session sequence using a heterogeneous graph to capture intricate behavior interactions and a hypergraph for contextual information integration. Specifically, we designed a heterogeneous enhancement module that deeply understands the intrinsic connections of behaviors and the interdependencies between different behavior types by enhancing the behavioral information of the central node, effectively capturing the complex dynamic interactions between nodes within the session to obtain accurate item embeddings. Concurrently, we propose a self-supervised training method for the module that mitigates location bias and minimizes the impact of noisy behaviors. For cross-session, we combine relevant contextual information through a hypergraph to achieve accurate recommendation results. Experimental results show that our proposed self-supervised enhancement method significantly improves recommendation performance and has a better performance compared to recommendation methods that only consider a single behavior.

Item Type:Article
Refereed:Yes
Divisions:Science > School of Mathematical, Physical and Computational Sciences > Department of Computer Science
ID Code:119793
Publisher:Institute of Electrical and Electronics Engineers (IEEE)

Downloads

Downloads per month over past year

University Staff: Request a correction | Centaur Editors: Update this record

Page navigation