TSGGuide: Recommendation Guide for Multivariate Time Series Generation

Beining Zhang, Kai Wu, Xiaoyu Zhang, Handing Wang, Jing Liu, Bei Wang

Keywords: time series generation, diffusion model, Channel Independence

Abstract: Multivariate Time Series Generation (MTSG) plays a crucial role in time series analysis, supporting tasks such as data augmentation and anomaly detection. While several methods exist for MTSG, recommending the most suitable method for new scenarios remains a significant challenge. Although prior work by (Ang et al., 2023a) provides guidance for selecting MTSG methods, it lacks coverage of recent diffusion-based methods and has limited exploration of channel-independent frameworks. We address these gaps by improving the recommendation guide, highlighting the effectiveness of a central discriminator within the channel-independent framework. Our revised guide makes three key recommendations: 1) VAE-based methods excel on small-scale datasets; 2) a channel-independent framework with the newly designed central discriminator is optimal in most cases; and 3) a diffusion-based method is preferable when ample data and computational resources are available.

Primary Area: learning on time series and dynamical systems

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 6449