CDC: Causal Domain Clustering for Multi-Domain Recommendation

📅 2025-07-09
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
In multi-domain recommendation, performance degradation arises from cross-domain discrepancies, while conventional domain grouping—based on business or data similarity—fails to capture genuine transfer relationships. Method: We propose Causal Domain Clustering (CDC), the first approach to integrate causal discovery into recommender systems for identifying inter-domain transferability. CDC constructs dual affinity matrices—Isolated-Domain and Mixed-Domain—to jointly model static structural associations and dynamic collaborative effects. It further introduces a cohesion-coefficient-based adaptive fusion mechanism, jointly optimizing domain clustering and source-domain selection. Contribution/Results: Evaluated across 50+ public and industrial domains, CDC achieves a 4.9% online eCPM lift, significantly outperforming traditional domain grouping and transfer learning strategies.

Technology Category

Application Category

📝 Abstract
Multi-domain recommendation leverages domain-general knowledge to improve recommendations across several domains. However, as platforms expand to dozens or hundreds of scenarios, training all domains in a unified model leads to performance degradation due to significant inter-domain differences. Existing domain grouping methods, based on business logic or data similarities, often fail to capture the true transfer relationships required for optimal grouping. To effectively cluster domains, we propose Causal Domain Clustering (CDC). CDC models domain transfer patterns within a large number of domains using two distinct effects: the Isolated Domain Affinity Matrix for modeling non-interactive domain transfers, and the Hybrid Domain Affinity Matrix for considering dynamic domain synergy or interference under joint training. To integrate these two transfer effects, we introduce causal discovery to calculate a cohesion-based coefficient that adaptively balances their contributions. A Co-Optimized Dynamic Clustering algorithm iteratively optimizes target domain clustering and source domain selection for training. CDC significantly enhances performance across over 50 domains on public datasets and in industrial settings, achieving a 4.9% increase in online eCPM. Code is available at https://github.com/Chrissie-Law/Causal-Domain-Clustering-for-Multi-Domain-Recommendation
Problem

Research questions and friction points this paper is trying to address.

Identifies inter-domain differences degrading multi-domain recommendation performance
Proposes causal clustering to model domain transfer and synergy effects
Optimizes domain grouping and training for improved recommendation accuracy
Innovation

Methods, ideas, or system contributions that make the work stand out.

Uses Isolated and Hybrid Domain Affinity Matrices
Introduces causal discovery for cohesion-based coefficient
Employs Co-Optimized Dynamic Clustering algorithm
🔎 Similar Papers
No similar papers found.
H
Huishi Luo
Institute of Artificial Intelligence, Beihang University, Beijing, China
Y
Yiqing Wu
Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China
Yiwen Chen
Yiwen Chen
Ph.D. student, S-Lab, Nanyang Technological University
Computer Vision3D GenerationGenerative Models
F
Fuzhen Zhuang
Institute of Artificial Intelligence, Beihang University, Beijing, China
D
Deqing Wang
SKLSDE, School of Computer Science, Beihang University, Beijing, China