Abstract:Sequential recommendation effectively addresses information overload by modeling users' temporal and sequential interaction patterns. To overcome the limitations of supervision signals, recent approaches have adopted self-supervised learning techniques in recommender systems. However, there are still two critical challenges that remain unsolved. Firstly, existing sequential models primarily focus on long-term modeling of individual interaction sequences, overlooking the valuable short-term collaborative relationships among the behaviors of different users. Secondly, real-world data often contain noise, particularly in users' short-term behaviors, which can arise from temporary intents or misclicks. Such noise negatively impacts the accuracy of both graph and sequence models, further complicating the modeling process. To address these challenges, we propose a novel framework called Self-Supervised Graph Neural Network (SelfGNN) for sequential recommendation. The SelfGNN framework encodes short-term graphs based on time intervals and utilizes Graph Neural Networks (GNNs) to learn short-term collaborative relationships. It captures long-term user and item representations at multiple granularity levels through interval fusion and dynamic behavior modeling. Importantly, our personalized self-augmented learning structure enhances model robustness by mitigating noise in short-term graphs based on long-term user interests and personal stability. Extensive experiments conducted on four real-world datasets demonstrate that SelfGNN outperforms various state-of-the-art baselines. Our model implementation codes are available at <a class="link-external link-https" href="https://github.com/HKUDS/SelfGNN" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on two aspects: 1. **Neglect of short - term collaborative relationships**: Existing sequential recommendation models mainly focus on the long - term behavior patterns of individual users and ignore the short - term collaborative relationships between different users. This neglect causes the model to be unable to fully utilize the immediate interaction information between users, thus affecting the accuracy of recommendations. 2. **Impact of data noise**: Data in the real world usually contains noise, especially the short - term behavior data of users. These noises may come from temporary intentions or misoperations. These noises will negatively affect the accuracy of graph models and sequence models and further complicate the modeling process. To solve these problems, the paper proposes a new framework named SelfGNN (Self - Supervised Graph Neural Network) for sequential recommendation. The main contributions of SelfGNN are as follows: - **Short - term collaborative graph encoding**: Divide the global user - item graph into multiple short - term graphs by time intervals and use graph neural networks (GNNs) to propagate collaborative high - order information. This not only captures the collaborative pattern but also includes short - term time information. - **Multi - level long - term sequence learning**: Conduct two - level sequence modeling to form complementary interests of different granularities. At the time - interval level, the features learned from different short - term graphs are processed into sequences to generate the long - term collaborative representation of users. At the instance level, model the complete behavior sequence of users through the self - attention mechanism. - **Personalized self - enhanced learning for denoising**: Design a self - enhanced learning task to correct the corresponding relationships in short - term graphs based on the long - term interests of users. This task learns the personalized weights of users and adjusts the noise level according to the interest stability of users, so that SelfGNN can identify short - term abnormal interactions and mitigate their impacts. Through these methods, SelfGNN aims to more accurately capture the dynamic interests of users and improve the robustness and accuracy of the recommendation system. Experimental results show that SelfGNN outperforms existing baseline models on multiple real - world datasets.

SelfGNN: Self-Supervised Graph Neural Networks for Sequential Recommendation

Self-Propagation Graph Neural Network for Recommendation

Adaptive self-supervised learning for sequential recommendation

Multi-dimensional Graph Neural Network for Sequential Recommendation

Graph convolutional network and self-attentive for sequential recommendation

Self-Attention Based Sequential Recommendation with Graph Convolutional Networks

Order-Aware Graph Neural Network for Sequential Recommendation

Attentive Sequential Model Based on Graph Neural Network for Next Poi Recommendation.

Category-aware Self-Supervised Graph Neural Network for Session-Based Recommendation

Dynamic Graph Neural Networks for Sequential Recommendation

Sequential Recommendation with Graph Neural Networks

Collaborative Sequential Recommendations via Multi-View GNN-Transformers

Graph Self-Attention Residual Connection Neural Network for Session-Based Recommendation

Self-supervised Graph Learning for Recommendation

RetaGNN: Relational Temporal Attentive Graph Neural Networks for Holistic Sequential Recommendation

A Time-Aware Graph Neural Network for Session-Based Recommendation

Position-enhanced and Time-aware Graph Convolutional Network for Sequential Recommendations

Self-supervised Graph Learning for Occasional Group Recommendation

Edge-Enhanced Global Disentangled Graph Neural Network for Sequential Recommendation

Multi-Grained Fusion Graph Neural Networks for Sequential Recommendation.

Self-Supervised Spatio-Temporal Graph Learning for Point-of-Interest Recommendation