Key Users Identification-Based Heterogeneous Hypergraph for Group Recommendation

One of the major challenges facing the group recommendation task is how to effectively aggregate member preferences to achieve optimal group consensus. Most of the traditional group recommendation methods use heuristic or attention-based strategies to integrate group preferences, however, these methods still suffer from the following two problems: 1) current neural network-based models ignore the fine-grained higher-order interactions among groups, users, and items; 2) the importance of identifying key users is ignored. To this end, this paper proposes a Key Users Identification-based Heterogeneous Hypergraph for Group Recommendation (UIGRec). We innovatively design two kinds of fine-grained graphs, namely disentangled nested hypergraph and key users heterogeneous graph. We construct a coarse-grained group-level preference graph, which aims to mine the multi-dimensional features of groups comprehensively and deeply. Then, we design a novel hypergraph convolutional network that more precisely captures the group’s higher-order interests. Finally, a method to identify key users across inter-groups is proposed, which accurately recognizes key users. The method effectively integrates the important influence of key users on group preferences. Experimental results show that the model proposed in this paper outperforms existing methods in terms of performance on multiple benchmark datasets.