基于因子分解机和图神经网络的推荐方法

Recommendation Method Based on Factor Decomposition Machine and Graph Neural Network

本文提出了一种融合因子分解机和图神经网络的推荐模型。通过递归地传播图中节点邻居的特征交叉并使用交叉特征关系域获取注意力权重。实验结果表明,图结构中节点的特征交叉能够提升推荐性能与可解释性,对后续研究具有借鉴意义。

This paper introduces a recommendation model combining factorization machines and graph neural networks.By recursively propagating cross-features among neighboring nodes and utilizing attention mechanisms,and use cross feature relationship domains to obtain attention weights.Experimental results validate its effectiveness,offering valuable insights for future research.