Bundle Recommendation with Item-level Causation-enhanced Multi-view Learning

TL;DR

BunCa tackles bundle recommendation by modeling asymmetric item-item influences within bundles using an item-level causation mechanism, integrated into a two-view framework (Coherent View and Cohesive View). The Coherent View employs a Multi-Prospect Causation Network to capture asymmetric item relations, while the Cohesive View uses LightGCN to propagate high-order user-bundle signals; both views are fused and reinforced with discrete and concrete contrastive learning. The approach yields state-of-the-art or competitive results on three benchmarks, notably excelling on iFashion where explicit anchor items emerge, and demonstrates the value of encoding asymmetric item relations for bundle construction and recommendation. Overall, BunCa provides a scalable, interpretable way to incorporate causation-aware item interactions into practical bundle recommendation systems.

Abstract

Bundle recommendation aims to enhance business profitability and user convenience by suggesting a set of interconnected items. In real-world scenarios, leveraging the impact of asymmetric item affiliations is crucial for effective bundle modeling and understanding user preferences. To address this, we present BunCa, a novel bundle recommendation approach employing item-level causation-enhanced multi-view learning. BunCa provides comprehensive representations of users and bundles through two views: the Coherent View, leveraging the Multi-Prospect Causation Network for causation-sensitive relations among items, and the Cohesive View, employing LightGCN for information propagation among users and bundles. Modeling user preferences and bundle construction combined from both views ensures rigorous cohesion in direct user-bundle interactions through the Cohesive View and captures explicit intents through the Coherent View. Simultaneously, the integration of concrete and discrete contrastive learning optimizes the consistency and self-discrimination of multi-view representations. Extensive experiments with BunCa on three benchmark datasets demonstrate the effectiveness of this novel research and validate our hypothesis.

Figures (6)

• Figure 1: Motivating examples of multi-prospect causation in bundle recommendation.
• Figure 2: The schematic illustration of our proposed model BunCa.
• Figure 3: Statistical distribution of High-level Influence Items within Bundles.
• Figure 4: Importance of asymmetric causation matrix $A$ on the performance of BunCa.
• Figure 5: The impact of key hyper-parameters on the performance of BunCa for iFashion. Figures (a) and (b) show the model performance calculated by $R@20$ and $N@20$ over the respective value ranges of $\beta$ and $L$.