Abstract:Advanced recommender systems usually involve multiple domains (such as scenarios or categories) for various marketing strategies, and users interact with them to satisfy diverse demands. The goal of multi-domain recommendation (MDR) is to improve the recommendation performance of all domains simultaneously. Conventional graph neural network based methods usually deal with each domain separately, or train a shared model to serve all domains. The former fails to leverage users' cross-domain behaviors, making the behavior sparseness issue a great obstacle. The latter learns shared user representation with respect to all domains, which neglects users' domain-specific preferences. In this paper we propose $\mathsf{H^3Trans}$, a hierarchical hypergraph network based correlative preference transfer framework for MDR, which represents multi-domain user-item interactions into a unified graph to help preference transfer. $\mathsf{H^3Trans}$ incorporates two hyperedge-based modules, namely dynamic item transfer (Hyper-I) and adaptive user aggregation (Hyper-U). Hyper-I extracts correlative information from multi-domain user-item feedbacks for eliminating domain discrepancy of item representations. Hyper-U aggregates users' scattered preferences in multiple domains and further exploits the high-order (not only pair-wise) connections to improve user representations. Experiments on both public and production datasets verify the superiority of $\mathsf{H^3Trans}$ for MDR.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to improve the recommendation performance in all domains in multi - domain recommendation systems. Specifically, when dealing with multi - domain recommendation, the existing graph neural network methods either model each domain separately, ignoring users' cross - domain behaviors; or construct a unified model to serve all domains, but this method ignores the specific characteristics of different domains, resulting in limited performance. The paper proposes a Hierarchical Hypergraph - based Network for Relevant Preference Transfer framework (H3Trans), aiming to improve the recommendation effects in all domains by integrating users' behavioral data in multiple domains and capture users' specific preferences in different domains at the same time. ### Main Contributions: 1. **Propose H3Trans**: A Hierarchical Hypergraph - based Network for Relevant Preference Transfer framework for multi - domain recommendation. As far as the authors know, this is the first study on hypergraph - based multi - domain recommendation preference transfer work. 2. **Dynamic Item Transfer Module (Hyper - I)**: By dynamically finding relevant items from user - item interactions in other domains and constructing hyperedges (hyperedge - i) of cross - domain item relationships, it helps to extract relevant preferences in cross - domain behaviors without introducing interfering information. 3. **Adaptive User Aggregation Module (Hyper - U)**: By connecting nodes of the same user in different domains, it aggregates users' scattered preferences in multiple domains and uses the attention mechanism to model high - order connections, thereby capturing the commonalities of multiple domains. 4. **Experimental Verification**: Extensive experiments have been carried out on large - scale production data sets and public data sets, and the results show the superiority of H3Trans in multi - domain recommendation. ### Model Architecture: - **Unified Multi - Domain Graph**: Integrate user and item nodes into a unified graph, where user nodes are represented separately in different domains, and item nodes are shared in all domains. - **Basic Graph Neural Network**: It includes an embedding module, a message passing module, a read - out module and a prediction module, which are used to learn node representations. - **Dynamic Item Transfer Module (Hyper - I)**: By constructing hyperedges to connect relevant items in different domains, it helps to capture relevant preferences in cross - domain behaviors. - **Adaptive User Aggregation Module (Hyper - U)**: By constructing hyperedges to connect nodes of the same user in different domains, it aggregates users' scattered preferences and uses the attention mechanism to model high - order connections. ### Mathematical Formulas: - **User - Item Interaction Matrix**: \[ R_m\in\mathbb{R}^{|U_m|\times|I_m|} \] where $r_{u,i}^m = 1$ indicates that user $u$ has interacted with item $i$ in domain $m$. - **Predicted Score**: \[ \hat{r}_{u,i}^m = f(z_u,z_i|D_m) \] where $z_u$ and $z_i$ represent the learned representations of user $u$ and item $i$ in domain $D_m$ respectively, and $f(\cdot)$ is a similarity function. Through these innovations, H3Trans can more effectively utilize users' behavioral data in multiple domains and improve the overall performance of the recommendation system.

Correlative Preference Transfer with Hierarchical Hypergraph Network for Multi-Domain Recommendation

HKGCL: Hierarchical graph contrastive learning for multi-domain recommendation over knowledge graph

Gmr-Rec: Graph Mutual Regularization Learning for Multi-Domain Recommendation

Sequential and Graphical Cross-Domain Recommendations with a Multi-View Hierarchical Transfer Gate

Heterogeneous and Clustering-Enhanced Personalized Preference Transfer for Cross-Domain Recommendation

Transfer Meets Hybrid: A Synthetic Approach for Cross-Domain Collaborative Filtering with Text

Adaptive Coordinators and Prompts on Heterogeneous Graphs for Cross-Domain Recommendations

HAMUR: Hyper Adapter for Multi-Domain Recommendation

Hierarchical Projection Enhanced Multi-behavior Recommendation.

Heterogeneous Graph-based Framework with Disentangled Representations Learning for Multi-target Cross Domain Recommendation

Multi-graph Convolutional Feature Transfer for Cross-domain Recommendation

Inter- and Intra-Domain Relation-Aware Heterogeneous Graph Convolutional Networks for Cross-Domain Recommendation.

Heterogeneous Multidomain Recommender System Through Adversarial Learning

Multi-Domain Recommendation to Attract Users via Domain Preference Modeling

Intra and Inter Domain HyperGraph Convolutional Network for Cross-Domain Recommendation

Graph Heterogeneous Multi-Relational Recommendation

CDR-ULPT:Cross-domain Recommendation for E-Commerce Using User-Level Preferences Transfer Network.

Embedding Transfer with Enhanced Correlation Modeling for Cross-Domain Recommendation.

Mining User Consistent and Robust Preference for Unified Cross Domain Recommendation

Multi-Head Attention and Knowledge Graph Based Dual Target Graph Collaborative Filtering Network

A Recommendation Method Based on Multi-Source Heterogeneous Hypergraphs and Contrastive Learning