Abstract:Fusion of multi-source information is one of the primary methods to alleviate data sparsity in recommender systems (RS). Hypergraphs have shown remarkable capabilities in dealing with the diversity of multi-source information, especially in modeling high-order user-item relationships. However, most hypergraph research focuses on constructing hypergraphs using a particular type of hyperedge, which might not fully capture the high-order implicit associations among heterogeneous information sources. Furthermore, existing Hypergraph Attention Networks (Hyper-GAT), mainly emphasizes on information propagation between nodes and hyperedges, insufficiently exploring density information within complex hyperedges. Moreover, when hypergraphs combine data from multiple heterogeneous graphs, redundant information from different viewpoints can reduce the effectiveness of multivariate modeling hyperedges. Thus, we propose a recommendation algorithm based on multi-source heterogeneous hypergraphs and contrastive learning (MHCLR), which improves recommendation accuracy through multi-source information fusion and higher-order information correlation. First, multi-source heterogeneous hypergraphs (MHC) are generated by combining distance, behavior, attribute, and prediction hyperedges. We mine associations among multi-source information, enhancing high-order semantic connections among users and items. Then, a Spatial Density Hypergraph Attention Network (SD-HGAT) is proposed based on composite hyperedges, which enriches the user and item embedding representations by focusing on nodes and hyperedges density. Finally, we design a multiple cross view contrastive learning (MCC) that compares views centered around knowledge graphs with hypergraphs, improving the accuracy of multivariate relationship modeling and enabling multi-level user profiling construction. It is observed that MHCLR outperforms the baselines in terms of Recall, Precision, and NDCG based on the experimental results by Yelp, Last-FM, and Douban.

Inductive Dummy-based Homogeneous Neighborhood Augmentation for Graph Collaborative Filtering.

Inductive Collaborative Filtering Via Relation Graph Learning

An Efficient Neighborhood-based Interaction Model for Recommendation on Heterogeneous Graph.

Hypergraph Contrastive Collaborative Filtering

GraphHINGE: Learning Interaction Models of Structured Neighborhood on Heterogeneous Information Network

Feature-enhanced Embedding Learning for Heterogeneous Collaborative Filtering

Hybrid Recommendation System with Graph Neural Collaborative Filtering and Local Self-attention Mechanism.

Recommendation on Heterogeneous Information Network with Type-Sensitive Sampling

Distribution-aware Hybrid Noise Augmentation in Graph Contrastive Learning for Recommendation

Personalized Recommendation Via Inductive Spatiotemporal Graph Neural Network

HGCH: A Hyperbolic Graph Convolution Network Model for Heterogeneous Collaborative Graph Recommendation

Graph DNA: Deep Neighborhood Aware Graph Encoding for Collaborative Filtering

Disentangled Hierarchical Attention Graph Neural Network for Recommendation

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

Community-Enhanced Contrastive Learning for Graph Collaborative Filtering

Self-Supervised Graph Attention Collaborative Filtering for Recommendation

Graph Collaborative Signals Denoising and Augmentation for Recommendation

GDSRec: Graph-Based DecentralizedCollaborative Filtering for SocialRecommendation

On Sampling Strategies for Neural Network-based Collaborative Filtering

An Effective Two-way Metapath Encoder over Heterogeneous Information Network for Recommendation

GDSRec: Graph-Based Decentralized Collaborative Filtering for Social Recommendation