... and neglect latent interaction patterns that are essential for improving recommendation performance. To address these limitations, we propose a novel model named hierarchical graph-based mixup for bundle recommendation ( HGMBR ), which integrates mixup techniques across both the item and bundle views to learn...