GAT-COBO

A PyTorch implementation for the paper below:
GAT-COBO: Cost-Sensitive Graph Neural Network for Telecom Fraud Detection.

Running GAT-COBO

To run the code, you need to have at least Python 3.7 or later versions.
1.In GAT-COBO/data directory，rununzip BUPT.zip and unzip Sichuan.zip to unzip the datasets;
2.Run python data_process.py to generate Sichuan and BUPT dataset in DGL;
3.Run python main.py to run GAT-COBO with default settings.
For other dataset and parameter settings, please refer to the arg parser in train.py. Our model supports both CPU and GPU mode.

Repo Structure

The repository is organized as follows:

• baselines/:code for all the baselines used in our paper;
• data/: dataset files;
• data_process.py: convert raw node features and adjacency matrix to DGL dataset;
• main.py: training and testing GAT-COBO;
• model.py: GAT-COBO model implementations;
• utils.py: utility functions for EarlyStopping,MixedDropout, MixedLinear, Cost matrix.

Running baselines

You can find the baselines in baselines directory. For example, you can run Player2Vec using:

python Player2Vec_main.py 

Citation

@article{hu2023gatcobo,
  title={GAT-COBO: Cost-Sensitive Graph Neural Network for Telecom Fraud Detection},
  author={Hu, Xinxin and Chen, Haotian and Zhang, Junjie and Chen, Hongchang and Liu, Shuxin and Li, Xing and Wang, Yahui and Xue, Xiangyang},
  journal={IEEE Transactions on Big Data},
  year={2023},
  doi={10.1109/TBDATA.2024.3352978}
}