LEX-GNN

January 12, 2025 ยท View on GitHub

PWC PWC

The author implementation of the CIKM 2024 short paper:
"LEX-GNN: Label-Exploring Graph Neural Network for Accurate Fraud Detection".
[Paper] [Poster]

Woochang Hyun, Insoo Lee, Bongwon Suh

Overview

Label-Exploring Graph Neural Network (LEX-GNN) is a GNN-based fraud detector that predicts the fraud probability of nodes in a semi-supervised manner and adaptively adjusts the message passing pipeline for improved accuracy.

Usage

  • Requirements: Python, Torch, and DGL
  • Dataset: Yelp and Amazon are loaded from dgl.data.fraud upon code execution.
  • Run: python main.py

Citation

@inproceedings{hyun2024lex,
  title={LEX-GNN: Label-Exploring Graph Neural Network for Accurate Fraud Detection},
  author={Hyun, Woochang and Lee, Insoo and Suh, Bongwon},
  booktitle={Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (CIKM'24)},
  year={2024}
}