Abstract:Multi-Behavior Recommendation (MBR) attracts much attention in recent years, whose goal is to improve the prediction of the target behavior (i.e., purchase) by exploiting multi-typed auxiliary behaviors (e.g., view, favorite and add-to-cart). Recently, leveraging Graph Convolutional Networks (GCNs) to capture collaborative signals has been the mainstream paradigm for MBR. However, the existing multi-behavior recommendation methods have suffered from two limitations. On the one hand, personalized user preferences hidden in multi-behavior data are not fully exploited. Users' multiple types of interactions, such as views, clicks, and so on, offer fine-grained and a deep understanding of the preferences. The importance of different types of behaviors should be carefully distinguished. On the other hand, these methods aggregate the original neighbors of target user and item independently. Users' preferences may change dynamically with a specific target item. Therefore, users' dynamic preferences based on specific items should be sufficiently considered. These limitations motivate us to propose a novel recommendation model DGAMR (Dual Graph Attention Networks for Multi-behavior Recommendation), which accurately learns user and item representation by multiple types of behaviors. First, we utilize node-level attention to learn the representation of users and items under specific behavior. Second, behavioral-level attention is used to aggregate different behaviors to generate the final representation of users and items. In addition, we learn the dynamic characteristics of target user and target item by modeling the dependency relation between them. Finally, we utilize the static and dynamic embedding of users/items to predict users' preferences for items. Extensive experimental results on two real-world datasets clearly show the effectiveness of our proposed model.

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

Enhanced Knowledge Transfer for Collaborative Filtering with Multi-Source Heterogeneous Feedbacks

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

Extracting latently overlapping users by graph neural network for non-overlapping cross-domain recommendation

Multi-Behavior Enhanced Heterogeneous Graph Convolutional Networks Recommendation Algorithm based on Feature-Interaction

Multi-modal Recommendation Based on Knowledge Graph

A Graphical and Attentional Framework for Dual-Target Cross-Domain Recommendation

KGCFRec: Improving Collaborative Filtering Recommendation with Knowledge Graph

Dual Attentive Graph Convolutional Networks for Cross-Domain Recommendation.

Deep Transfer Collaborative Filtering with Geometric Structure Preservation for Cross-Domain Recommendation

Dual Channel Hypergraph Collaborative Filtering

Graph Heterogeneous Multi-Relational Recommendation

Dual Graph Attention Networks for Multi-Behavior Recommendation

Preference-aware Graph Attention Networks for Cross-Domain Recommendations with Collaborative Knowledge Graph

Knowledge-aware Multi-view Cross Learning for Edge-based Collaborative Recommendation

Cross Domain Recommendation Via Bi-directional Transfer Graph Collaborative Filtering Networks.

Multi-task Feature Learning for Social Recommendation

Dual-Channel Feature Enhanced Collaborative Filtering Recommendation Algorithm.

Graph Collaborative Filtering Based on Dual-Message Propagation Mechanism.

Neighborhood Aggregation Collaborative Filtering Based on Knowledge Graph

Broad Learning based Multi-Source Collaborative Recommendation.