Multi-interaction fusion collaborative filtering for social recommendation

作者：

Highlights：

• We focus on multi-graph aggregation and divide data into four graphs.

• To capture influence of user interests, a dynamic attention module is designed.

• A mutualistic attention mechanism is designed to merge and transmit factors.

• We design a neural collaborative filtering module to fuse the characteristics.

摘要

•We focus on multi-graph aggregation and divide data into four graphs.•To capture influence of user interests, a dynamic attention module is designed.•A mutualistic attention mechanism is designed to merge and transmit factors.•We design a neural collaborative filtering module to fuse the characteristics.

论文关键词：Social recommendation,Graph neural network,Heterogeneous network,Mutualistic attention mechanism

论文评审过程：Received 5 July 2021, Revised 9 May 2022, Accepted 15 May 2022, Available online 26 May 2022, Version of Record 31 May 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117610