Diffusion models have been recently used for anomaly detection (AD) in images. In this paper we investigate whether they can also be leveraged for AD on multivariate time series (MTS). We test two diffusion-based models and compare them to several strong neural baselines. We also extend the PA%K protocol, by computing a ROCK-AUC metric, which is agnostic to both the detection threshold and the ratio K of correctly detected points. Our models outperform the baselines on synthetic datasets and are competitive on real-world datasets, illustrating the potential of diffusion-based methods for AD in multivariate time series.
@article{arxiv.2311.01452,
title = {Time Series Anomaly Detection using Diffusion-based Models},
author = {Ioana Pintilie and Andrei Manolache and Florin Brad},
journal= {arXiv preprint arXiv:2311.01452},
year = {2023}
}
Comments
Accepted at the AI4TS workshop of the 23rd IEEE International Conference on Data Mining (ICDM 2023), 9 pages, 7 figures, 2 tables