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

Researchers have unveiled EpicCBR, a new AI framework for recommending product bundles to users who have no prior interaction history. It uses multi-view contrastive learning to analyze item relations and user profiles, outperforming existing state-of-the-art models by up to 387% in cold-start scenarios. The model was validated on three popular benchmarks and its code is publicly available on GitHub, promising major improvements for e-commerce and streaming platforms.