Author(s)

Danhao Ma, Lichen Zhang

Corresponding Author

Danhao Ma

ABSTRACT

With the rapid development of social networks and online platforms, social recommendation has become an important task in personalized recommendation systems. Social relationships have a significant impact on users' decisions, and recommendations from friends play a substantial role in whether a user purchases a particular item. Users' browsing, bookmarking, adding to cart, and purchase behaviors contain a wealth of information. Traditional recommendation models only focus on either users' social relationships or individual behaviors without connecting them. In this paper, we introduce a social recommendation model called MBSR, which utilizes graph convolution to extract users' social relationship lists and constructs multiple hypergraphs by combining a user's behavioral sequences. Different hypergraphs reflect users' different behavior patterns towards various items, ultimately predicting user ratings for items. Experimental results on a dataset demonstrate that the proposed model effectively improves recommendation performance.

KEYWORDS

Recommendation algorithm, Graph convolution, Hypergraph, Deep learning