The official implemantation of AAAI-24 paper DiG-In-GNN: Discriminative Feature Guided GNN-Based Fraud Detector against Inconsistencies in Multi-Relation Fraud Graph. Paper Link
The folder Code4DiGInGNN in this repo contains the code version we submitted to AAAI as the supplementary file, and there is a README.md included in it. This version runs slowly on T-Finance dataset because it uses many for loops in the neighbor selector, instead of faster tensor operations that can be accelerated by parallel computing on cuda devices. We have written a faster version and will share it after the author is through with their busy graduation season.
The codes are based on PC-GCN.
If you have any problems or trouble, you can write an issue to let us know.
