tsfpaper is a curated reading list of academic papers on time series forecasting and spatio-temporal prediction methods.
The repository addresses the challenge of navigating the growing body of research in time series and spatio-temporal forecasting by organizing papers according to the types of models they employ. Rather than presenting an exhaustive bibliography, it structures papers by model category, helping researchers and practitioners quickly locate work relevant to their specific methodological interests, whether in recurrent neural networks, temporal convolutional networks, transformers, or other deep learning approaches.
This resource suits researchers exploring the landscape of forecasting methods, practitioners evaluating different modeling approaches for their applications, and students building foundational knowledge in time series analysis. The categorical organization by model type makes it particularly useful for those investigating how specific architectures perform on forecasting tasks or comparing methodological approaches. The inclusion of both time series forecasting and spatio-temporal prediction papers means it covers both univariate and multivariate scenarios with spatial dependencies.
The project shows consistent engagement with its reading list, reflecting ongoing curation as new research emerges in the field. The repository maintains active organization of papers across multiple model categories, indicating that the maintainer regularly updates the collection to reflect developments in deep learning approaches to forecasting.
