Graphical contrastive learning for multi-interest sequential recommendation

Shunpan Liang, Qianjin Kong, Yu Lei, Chen Li

Abstract: Highlights•GCL4MI integrates GNN and CL to model user interest and address data sparsity issues.•We explore CL sampling strategies with GNN, a novel approach for multi-interest models.•Extensive experiments show GCL4MI outperforms state-of-the-arts on recall, NDCG, and HR.