EpicCBR: Item-Relation-Enhanced Dual-Scenario Contrastive Learning for Cold-Start Bundle Recommendation

TL;DR

EpicCBR tackles cold-start bundle recommendation by introducing a dual-scenario, multi-view contrastive framework that jointly leverages item-pair relations and popularity-aware embeddings. It mines four item-pair relation types from user-item and bundle-item graphs, injects cross-domain signals to enrich item representations, and uses a popularity-based mechanism to construct robust bundle representations for unseen bundles. The model integrates cold-start and warm-start paths through a scenario-weighted fusion and optimizes with a combination of BPR loss, dual-scenario contrastive losses, and item-pair losses, achieving superior performance on three real-world datasets, including a remarkable 387% Recall@20 gain on iFashion. This work demonstrates that explicit item-level relations and popularity cues, combined with dual-scenario contrastive learning, can significantly improve robustness and accuracy in cold-start bundle recommendations, with potential extensions to other recommendation domains.

Abstract

Bundle recommendation aims to recommend a set of items to users for overall consumption. Existing bundle recommendation models primarily depend on observed user-bundle interactions, limiting exploration of newly-emerged bundles that are constantly created. It pose a critical representation challenge for current bundle methods, as they usually treat each bundle as an independent instance, while neglecting to fully leverage the user-item (UI) and bundle-item (BI) relations over popular items. To alleviate it, in this paper we propose a multi-view contrastive learning framework for cold-start bundle recommendation, named EpicCBR. Specifically, it precisely mine and utilize the item relations to construct user profiles, identifying users likely to engage with bundles. Additionally, a popularity-based method that characterizes the features of new bundles through historical bundle information and user preferences is proposed. To build a framework that demonstrates robustness in both cold-start and warm-start scenarios, a multi-view graph contrastive learning framework capable of integrating these diverse scenarios is introduced to ensure the model's generalization capability. Extensive experiments conducted on three popular benchmarks showed that EpicCBR outperforms state-of-the-art by a large margin (up to 387%), sufficiently demonstrating the superiority of the proposed method in cold-start scenario. The code and dataset can be found in the GitHub repository: https://github.com/alexlovecoding/EpicCBR.

Figures (3)

• Figure 1: The overall framework of EpicCBR. (1) Multi-view Representation Learning comprises a Warm-Scenario Module (User-Bundle View, User-Item View, Bundle-Item View) for historical interaction modeling and a Cold-Scenario Module (User-Item-Bundle View, Bundle-Item-User View) for deducing potential interactions. (2) Fusion & Prediction integrates representations from both scenarios via contrastive losses to generate recommendation scores.
• Figure 2: Visualizing Item-Pair Relations
• Figure 3: Experiments for scenario weight on dataset all