Causal Deconfounding via Confounder Disentanglement for Dual-Target Cross-Domain Recommendation

2024-04-17

arXiv.org

AI Summary

In dual-objective cross-domain recommendation (CDR), confounding factors—such as free shipping and promotions—distort the modeling of users’ true preferences, while both intra-domain and cross-domain confounding effects are often overlooked. To address this, we propose CD²CDR, a causal deconfounding framework: (1) it is the first to jointly model and disentangle intra- and cross-domain confounders in dual-objective CDR; (2) it introduces a causal deconfounding module leveraging backdoor adjustment to preserve the positive influence of confounders on interaction occurrence while eliminating their negative bias on user preference representation; and (3) it integrates graph neural networks with domain-adaptive representation learning to enhance transfer robustness. Evaluated on seven real-world datasets, CD²CDR achieves average improvements of 12.6% and 18.3% in recommendation accuracy on the primary and auxiliary domains, respectively—significantly outperforming state-of-the-art methods—and demonstrates the critical role of causal modeling in improving generalization and robustness in CDR.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

In recent years, dual-target Cross-Domain Recommendation (CDR) has been proposed to capture comprehensive user preferences in order to ultimately enhance the recommendation accuracy in both data-richer and data-sparser domains simultaneously. However, in addition to users' true preferences, the user-item interactions might also be affected by confounders (e.g., free shipping, sales promotion). As a result, dual-target CDR has to meet two challenges: (1) how to effectively decouple observed confounders, including single-domain confounders and cross-domain confounders, and (2) how to preserve the positive effects of observed confounders on predicted interactions, while eliminating their negative effects on capturing comprehensive user preferences. To address the above two challenges, we propose a Causal Deconfounding framework via Confounder Disentanglement for dual-target Cross-Domain Recommendation, called CD2CDR. In CD2CDR, we first propose a confounder disentanglement module to effectively decouple observed single-domain and cross-domain confounders. We then propose a causal deconfounding module to preserve the positive effects of such observed confounders and eliminate their negative effects via backdoor adjustment, thereby enhancing the recommendation accuracy in each domain. Extensive experiments conducted on seven real-world datasets demonstrate that CD2CDR significantly outperforms the state-of-the-art methods.

Problem

• Decouple single-domain and cross-domain confounders in CDR
• Preserve positive effects of confounders on recommendations
• Eliminate negative effects of confounders on user preferences

Innovation

• Disentangles single and cross-domain confounders effectively
• Uses causal deconfounding to preserve positive confounder effects
• Applies backdoor adjustment to eliminate negative confounder impacts