Error-Bounded Graph Anomaly Loss for GNNs

This repository contains the code package for the TNNLS paper:

A Synergistic Approach for Graph Anomaly Detection with Pattern Mining and Feature Learning.

Authors: Tong Zhao ([email protected]), Tianwen Jiang, Neil Shah, and Meng Jiang.

and the CIKM'20 paper:

Error-Bounded Graph Anomaly Loss for GNNs.

Authors: Tong Zhao ([email protected]), Chuchen Deng, Kaifeng Yu, Tianwen Jiang, Daheng Wang, and Meng Jiang.

Usage

1. Dependencies

This code package was developed with Python 3.6.8 and PyTorch 1.0.1.post2. A detailed dependencies list can be found in requirements.txt and can be installed by:

pip install -r requirements.txt

2. Data

Data files are located at /data/[dataset]/, a simple example of loading the data can be found here. Specifically, [dataset]_graph_u2p.pkl is the pickled sparse adjacency matrix (csr_matrix) and [dataset]_labels_u.pkl is the pickled user labels.

3. Run

To train the model, run

python -m src.main

list of arguments can be found at here.

Cite

If you find this repository useful in your research, please cite our papers:

@ARTICLE{zhao2021synergistic,
  author={Zhao, Tong and Jiang, Tianwen and Shah, Neil and Jiang, Meng},
  journal={IEEE Transactions on Neural Networks and Learning Systems}, 
  title={A Synergistic Approach for Graph Anomaly Detection With Pattern Mining and Feature Learning}, 
  year={2021},
  volume={33},
  number={6},
  pages={2393-2405},
  doi={10.1109/TNNLS.2021.3102609}}

@inproceedings{zhao2020error,
  title={Error-Bounded Graph Anomaly Loss for GNNs},
  author={Zhao, Tong and Deng, Chuchen and Yu, Kaifeng and Jiang, Tianwen and Wang, Daheng and Jiang, Meng},
  booktitle={Proceedings of the 29th ACM International Conference on Information \& Knowledge Management},
  pages={1873--1882},
  year={2020}
}