Regularized Framework on Heterogeneous Hypergraph Model for Personal Recommendation
Tingting Zhu,Jianrui Chen,Zhihui Wang,Di Wu
DOI: https://doi.org/10.1007/978-981-19-8152-4_11
2022-01-01
Abstract:Filtering-based recommendations are one of the most widely used recommendation algorithms in recent years. Most of which are based on simple graph to construct network model, and the nodes connected by edges are all pairwise relationships. In practice, some relationships are more complex than pairwise relationships. Moreover, collaborative filtering only focused on the relationships between users and users, items and items, users and items, without considering the relations between items and tags. To address these problems and improve the accuracy of the recommendation algorithms, we propose a regularized framework based on heterogeneous hypergraph, which integrates tag information into the recommendation system. Firstly, the hyperedges are built for each user and all items those are rated by the user, and then the similarity index between items in the hyperedge is calculated. Secondly, the relational graph between items and tags is constructed. Thirdly, we establish a regularization framework, and minimize the cost function for scoring prediction and recommendation. Finally, we verify the effectiveness of our proposed algorithm on Movielens-100k, Restaurant & consumer and Filmtrust datasets, and the diverse simulation results show that our proposed algorithm gains better recommendation performance.