An edge feature aware heterogeneous graph neural network model to support tax evasion detection

Bin Shi, Bo Dong, Yiming Xu, Jiaxiang Wang, Yunfan Wang, Qinghua Zheng

Abstract: Highlights•Graph neural network will help improve the tax evasion detection performance.•Comprehensive features could be extracted by metapaths and hierarchical attention.•Our method has been developed and applied in cooperation with Chinese Revenue Agency.•Experiments on real-world dataset shows our method outperforms over 10+ baselines.