Revisiting Graph Contrastive Learning on Anomaly Detection from a Structural Imbalance Perspective

Code structure

folder	description
Data	Datasets.
AD-GCL	AD-GCL implementation code is provided.

Datasets

Dataset	# Nodes	# Edges	# Attribute	# Anomalies
Cora	2,708	5,429	1,433	5.5%
Citeseer	3,327	4,732	3,703	4.5%
Pubmed	19,717	44,338	500	3.0%
Bitcoinotc	5,881	35,592	128	5.1%
BITotc	4,863	28,473	128	6.2%
BITalpha	3,219	19,364	128	9.3%

Usage

# Cora
python ./AD-GCL/run.py --dataset 'cora' --lr 5e-3 --num_epoch 200 --threshold 7 --gpu_id 0

# Citeseer
python ./AD-GCL/run.py --dataset 'citeseer' --lr 3e-3 --num_epoch 200 --threshold 6 --gpu_id 0

# Pubmed
python ./AD-GCL/run.py --dataset 'pubmed' --lr 4e-3 --num_epoch 100 --threshold 8 --gpu_id 0

# Bitcoinotc
python ./AD-GCL/run.py --dataset 'bitcoinotc' --lr 4e-4 --num_epoch 100 --threshold 8 --gpu_id 0

# BITotc
python ./AD-GCL/run.py --dataset 'bitotc' --lr 5e-4 --num_epoch 100 --threshold 7 --gpu_id 0

# BITalpha
python ./AD-GCL/run.py --dataset 'bitalpha' --lr 5e-3 --num_epoch 100 --threshold 8 --gpu_id 0

Dependencies

• Python 3.8.13
• PyTorch 1.12.1
• dgl 0.4.3.post1
• Scipy 1.9.1
• Tqdm 4.64.1

Reference

If you find this repository useful in your research, please consider citing the following paper:

@inproceedings{xu2025revisiting,
  title={Revisiting Graph Contrastive Learning on Anomaly Detection: A Structural Imbalance Perspective},
  author={Xu, Yiming and Peng, Zhen and Shi, Bin and Hua, Xu and Dong, Bo and Wang, Song and Chen, Chen},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={39},
  number={12},
  pages={12972--12980},
  year={2025}
}