Abstract: Bundle recommendation, which aims to suggest personalized groups of items to users, has emerged as an effective strategy in modern e-commerce platforms, allowing users to satisfy diverse needs with a single purchase. Existing models based on graph convolution networks face limitations in capturing the complex relationships between users, items, and bundles as well as effectively integrating information from different interaction views. To address these challenges, HC4BR(Hypergraph Convolution for Bundle Recommendation), a novel transitional hypergraph convolutional model for bundle recommendation is proposed. It employs transitional hypergraph convolution to capture intricate and higher-order relationships among entities, integrates different views based on various user preference strategies during information propagation and incorporates contrastive learning to refine the model′s ability to distinguish between similar entities, improving its capacity to explore subtle distinctions while reinforcing meaningful connections. Experimental results on three public datasets demonstrate that the model outperforms the state-of-the-art method by 3.39%, 2.00% and 14.70% respectively, showcasing its effectiveness in addressing the complex challenges of bundle recommendation and its potential to enhance user experience and business outcomes in e-commerce platforms.