Transition Information Enhanced Disentangled Graph Neural Networks for session-based recommendation

作者：

Highlights：

• We emphasize the importance of modeling fine-grained transitions between items.

• We are the first to consider fine-grained transition information modeling.

• We leverage the position information to capture the transitions between items.

• We represent the item with disentangled representations of factors.

• Experimental results show that our proposed model outperforms the SOAT methods.

摘要

•We emphasize the importance of modeling fine-grained transitions between items.•We are the first to consider fine-grained transition information modeling.•We leverage the position information to capture the transitions between items.•We represent the item with disentangled representations of factors.•Experimental results show that our proposed model outperforms the SOAT methods.

论文关键词：Session-based recommendation,Graph neural networks,Disentangled representation learning,Contrastive learning.

论文评审过程：Received 8 April 2022, Revised 7 July 2022, Accepted 30 July 2022, Available online 6 August 2022, Version of Record 20 August 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118336