Travel recommendation via diffusion-guided multiplex graph contrastive learning.

Travel recommendation (TR) demonstrates immense commercial potential in global economic development. However, previous research has relied solely on single-form interaction attributes between users and travel products, overlooking the impact of different interaction attributes on the final travel product and the distinctions among these attributes. Additionally, travel data inherently suffers from data sparsity. To bridge this gap, this paper proposes a novel diffusion-guided multiplex graph contrastive learning (CL) for TR framework, namely DMGCLTR. It leverages an enhanced Diffusion Model (DM) to capture global interaction patterns across diverse attributes between users and travel products. Through a designed attribute-chain representation learner and perceptual encoder, it enables granular discovery of how different interaction attributes influence and depend upon the final selected travel product. Furthermore, we adopt a contrastive learning approach to jointly optimize these tasks and enhance its symmetry. Extensive experiments across three real-world datasets demonstrate DMGCLTR's superior performance. On average, it outperforms the best baseline by 5.20% in HR@5 and 7.01% in MRR across all datasets. This enhanced TR method offers a novel perspective for innovation and integrated development in the smart tourism industry.