A social recommendation model based on adaptive residual graph convolution networks
Incorporating social information in the recommendation algorithm based on graph neural network (GNN) alleviates the data sparsity and cold-start problems to a certain extent, and effectively improves the recommendation performance of the model. However, there are still shortcomings in the existing s...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
PeerJ Inc.
2025-07-01
|
| Series: | PeerJ Computer Science |
| Subjects: | |
| Online Access: | https://peerj.com/articles/cs-3010.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Incorporating social information in the recommendation algorithm based on graph neural network (GNN) alleviates the data sparsity and cold-start problems to a certain extent, and effectively improves the recommendation performance of the model. However, there are still shortcomings in the existing studies: on the one hand, the potential effect of noise in the raw data is ignored; on the other hand, only relying on the single interaction information between the user and the item and failing to make full use of the rich multi-aided information. These factors lead to an unsatisfactory learning effect of the model. To address the above problems, we propose a social recommendation model based on adaptive residual graph convolutional networks (SocialGCNRI). Specifically, we use the idea of fast Fourier transform (FFT), a filtering algorithm in the field of signal processing, to attenuate the raw data noise in the frequency domain, followed by utilizing the user-social relations, item-association relations, and user-item-interaction relations to form a heterogeneous graph to supplement the model information, and finally using a graph convolution algorithm with an adaptive residual graph to improve the expressive power of the model. Extensive experiments on two real datasets show that SocialGCNRI outperforms state-of-the-art social recommendation methods on a variety of common evaluation metrics. |
|---|---|
| ISSN: | 2376-5992 |