Group Recommendation Method for Hypergraph Message Passing Meta-Path Mask

Focusing on addressing the challenge of low recommendation accuracy within group recommendation contexts, a Group recommendation method for Hypergraph Message passing Meta-path masks (MMGH) is proposed. Initially, a hypergraph model is employed to capture the intricate higher-order interactions among users and groups, enabling the aggregation of intra-group dependencies. Then, a randomized discard-based message delivery method is applied to the GNN for hypergraph message delivery this allows the same node to propagate different messages to different group members. Then, the rich interaction information of users between groups is used to construct hyperedge through intergroup similarity, and user personal information to construct the meta-path aggregation model with random discard; finally, the group consensus information from two different views is modeled. The effectiveness of the proposed method is validated through experiments conducted on real datasets.