Collaborative filtering-based recommender systems that rely on a single type of behavior often encounter serious sparsity issues in real-world applications, leading to unsatisfactory performance. Multi-behavior Recommendation (MBR) is a method that seeks to learn user preferences, represented as vector embeddings, from auxiliary information. By leveraging these preferences for target behavior recommendations, MBR addresses the sparsity problem and improves the accuracy of recommendations. In this paper, we propose MB-HGCN, a novel multi-behavior recommendation model that uses a hierarchical graph convolutional network to learn user and item embeddings from coarse-grained on the global level to fine-grained on the behavior-specific level. Our model learns global embeddings from a unified homogeneous graph constructed by the interactions of all behaviors, which are then used as initialized embeddings for behavior-specific embedding learning in each behavior graph. We also emphasize the distinct of the user and item behaviorspecific embeddings and design two simple-yet-effective strategies to aggregate the behavior-specific embeddings for users and items, respectively. Finally, we adopt multi-task learning for optimization. Extensive experimental results on three real-world datasets demonstrate that our model significantly outperforms the baselines, achieving a relative improvement of 73.93% and 74.21% for HR@10 and NDCG@10, respectively, on the Tmall datasets.

翻译：基于单一行为类型的协同过滤推荐系统在实际应用中常面临严重的数据稀疏问题，导致性能不佳。多行为推荐（MBR）是一种从辅助信息中学习用户偏好（以向量嵌入表示）的方法。通过利用这些偏好进行目标行为推荐，MBR能缓解稀疏性问题并提升推荐准确性。本文提出MB-HGCN——一种新颖的多行为推荐模型，采用分层图卷积网络，从全局层面的粗粒度学习到行为特定层面的细粒度学习，逐步学习用户和物品嵌入。该模型首先从所有行为交互构建的统一同构图中学全局嵌入，随后将其作为初始化嵌入用于每个行为图中的行为特定嵌入学习。我们同时强调用户与物品行为特定嵌入的差异性，并设计两种简洁高效的策略分别聚合用户和物品的行为特定嵌入。最终采用多任务学习进行优化。在三个真实数据集上的大量实验结果表明，本模型显著优于基线方法：在天猫数据集的HR@10和NDCG@10指标上分别实现73.93%和74.21%的相对提升。