This is an official implentation of DiffAT: Effective data augmentation with diffusion models for time series forecasting.
DiffAT is a diffusion-based data augmentation framework for time-series forecasting.
Due to current deadline time constraints, this source code is not yet a complete pipeline project, we will supplement, improve, and revise the readme file, shell file and source code by the end of 2025.
You can obtain all the eight benchmarks from Google Drive provided in Autoformer. All the datasets are well pre-processed and can be used easily.