Interdependency Matters: Graph Alignment for Multivariate Time Series Anomaly Detection (ICDM 2024)

This repository provides a PyTorch implementation of MADGA, which transforms the unsupervised anomaly detection to graph alignment problem.

Framework

Data

We test our method for five publicly processed datasets, e.g., SWaT, WADI, PSM, MSL, and SMD.

• SWaT
• WADI
• PSM is released in RANSynCoders.
• MSL is released in GDN.
• SMD is released in OmniAnomaly.

mkdir Dataset
cd Dataset
mkdir input

Download the dataset in Data/input.

Train

• train for MADGA For example, training for WADI

sh script/run_WADI.sh

• train for DeepSVDD, DeepSAD, DROCC, and ALOCC.

python3 baseline_train.py --name SWaT --model DeepSVDD

• train for USAD and DAGMM We report the results by the implementations in the following links: USAD and DAGMM

Test

We provide the pretained model of MADGA.

For example, testing for WADI

sh script/test_WADI.sh

BibTex Citation

If you find this paper or repository helpful, please cite our paper. Thanks a lot~

@article{wang2024interdependency,
  title={Interdependency Matters: Graph Alignment for Multivariate Time Series Anomaly Detection},
  author={Wang, Yuanyi and Sun, Haifeng and Wang, Chengsen and Zhu, Mengde and Wang, Jingyu and Tang, Wei and Qi, Qi and Zhuang, Zirui and Liao, Jianxin},
  journal={arXiv preprint arXiv:2410.08877},
  year={2024}
}