A graph neural network-based node classification model on class-imbalanced graph data

作者：

Highlights：

• A novel GNN-based imbalanced node classification model GNN-INCM is proposed.

• Two cooperative modules are given to accurately learn robust node embeddings for both majority and minority classes.

• A sample strategy is designed to ensure that the embeddings of hard nodes are correctly represented.

• To improve overall performance, a hard sample-based KD method is introduced to jointly train multiple GNN-INCM models.

摘要

•A novel GNN-based imbalanced node classification model GNN-INCM is proposed.•Two cooperative modules are given to accurately learn robust node embeddings for both majority and minority classes.•A sample strategy is designed to ensure that the embeddings of hard nodes are correctly represented.•To improve overall performance, a hard sample-based KD method is introduced to jointly train multiple GNN-INCM models.

论文关键词：Graph convolutional network,Class-imbalanced,Node classification,Knowledge distillation,Hard sample

论文评审过程：Received 28 July 2021, Revised 2 March 2022, Accepted 3 March 2022, Available online 9 March 2022, Version of Record 19 March 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108538