STIM

This repository is the official implementation of "Address Anomalies at Critical Crossroads for Graph Anomaly Detection", accepted by TKDE.

Overview

Our implementation for STIM is based on PyTorch.

Requirments

This code requires the following:

• Python==3.8
• PyTorch==1.9.0+cu111
• Pytorch Geometric==2.3.0
• Numpy==1.21.2
• Scipy==1.9.3
• Scikit-learn==1.1.2
• NetworkX==2.8.8
• OGB==1.3.5
• DGL==0.4.3
• DGL-cu111==0.6.1 (Do not use the version which is newer than that!)

Usage

Step1: Pre-processing

python preprocessing.py

Step2: Anomaly Detection

python run.py

Baselines

All baselines and their URLs are as follows:

• ANOMALOUS [Paper] [Code]
• DGI [Paper] [Code]
• CoLA [Paper] [Code]
• ANEMONE [Paper] [Code]
• SL-GAD [Paper] [Code]
• Sub-CR [Paper] [Code]
• GRADATE [Paper] [Code]
• TAM [Paper] [Code]

Cite

If you compare with, build on, or use aspects of this work, please cite the following:

@article{yan2025address,  
  title={Address Anomalies at Critical Crossroads for Graph Anomaly Detection},  
  author={Yan, Junyi and Zuo, Enguang and Liang, Ke and Liu, Meng and Li, Miaomiao and Liu, Xinwang and Lv, Xiaoyi and Lu, Kai},  
  journal={IEEE Transactions on Knowledge and Data Engineering},  
  year={2025},  
  publisher={IEEE}  
}