Heterogeneous Sequel-Aware Graph Neural Networks for Sequential Learning

2025-06-05

AI Summary

This work addresses the sequence recommendation task, where conventional graph neural networks (GNNs) neglect temporal dependencies among items. We propose a dynamic collaborative-aware model that jointly encodes item-level temporal succession patterns and high-order heterogeneous user–item graph structures. Our key contribution is the first explicit integration of item succession relations—captured as time-aware sequential transitions—into a heterogeneous GNN framework, enabling joint modeling of sequential semantics and structural graph representations. To enhance expressiveness, we introduce neighbor-aware dynamic aggregation and multi-source sequence-augmented embedding mechanisms. Extensive experiments on three real-world and synthetic benchmark datasets demonstrate that our method consistently outperforms strong baselines—including Transformers, standard GNNs, and autoencoders—achieving up to a 12.7% improvement in Recall@10. These results empirically validate the effectiveness of explicit temporal succession modeling in boosting recommendation accuracy.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Graph-based recommendation systems use higher-order user and item embeddings for next-item predictions. Dynamically adding collaborative signals from neighbors helps to use similar users' preferences during learning. While item-item correlations and their impact on recommendations have been studied, the efficacy of temporal item sequences for recommendations is much less explored. In this paper, we examine temporal item sequence (sequel-aware) embeddings along with higher-order user embeddings and show that sequel-aware Graph Neural Networks have better (or comparable) recommendation performance than graph-based recommendation systems that do not consider sequel information. Extensive empirical results comparing Heterogeneous Sequel-aware Graph Neural Networks (HSAL-GNNs) to other algorithms for sequential learning (such as transformers, graph neural networks, auto-encoders) are presented on three synthetic and three real-world datasets. Our results indicate that the incorporation of sequence information from items greatly enhances recommendations.

Problem

• Explores efficacy of temporal item sequences for recommendations
• Compares sequel-aware GNNs with non-sequel graph-based systems
• Demonstrates sequence information enhances recommendation performance

Innovation

• Heterogeneous sequel-aware GNNs for sequential learning
• Dynamic collaborative signals from neighbor embeddings
• Temporal item sequence enhances recommendation performance