基于元图注意力网络的异质信息网络链路预测方法

Meta Graph Attention Network Based Link Prediction Method in Heterogeneous Information Networks

由于元路径存在节点之间语义信息模糊、联通性不强的缺陷,现有的基于元路径的异质信息网络链路预测方法不足以很好地捕捉复杂的结构信息和丰富的语义信息,从而影响链路预测性能的提升。为解决此问题,提出基于元图注意力网络的异质信息网络链路预测方法,综合利用元图内节点层面和元图间语义层面两种注意力机制学习节点向量表示,进而用学习到的节点向量表示进行链路预测。真实世界数据集上的实验结果表明,与最新的基准方法相比,所提方法可显著提高异质信息网络链路预测性能。

Due to the fact that metapaths have the defects of fuzzy semantic information and weak connectivity between nodes,existing metapaths based link prediction methods in heterogeneous information networks still fall short of capturing complex structure and rich semantic information,which has side effects on performance improvements of link prediction.To address this problem,the meta graph attention network based link prediction method in heterogeneous information networks(MGAT-LP)is proposed.Node embeddings are comprehensively learned with both intra-metagraph node-level and inter-metagraph semantic-level attention mechanism.Then the learned node embeddings are used for link prediction.Experimental results on real-world datasets show that,compared with state-of-the-art baseline methods,the proposed method can significantly improve the performance of link prediction in heterogeneous information networks.