This repository serves as an extensive, curated reading list for time series forecasting (TSF) and spatio-temporal forecasting (STF). It categorizes over 400 research papers, primarily by model type, targeting researchers and practitioners in machine learning and data science. The benefit is a structured overview of the state-of-the-art, including recent arXiv preprints and influential conference/journal publications.

How It Works

The repository organizes papers into categories such as Transformers, RNNs, MLPs, GNNs, State Space Models (SSMs), Diffusion Models, LLMs, and more. Each entry includes publication details, a link to the paper, and often a link to associated code. The categorization is based on the primary architectural approach described in the paper, aiding users in navigating specific modeling paradigms.

Quick Start & Requirements

This is a curated list of research papers, not a software library. No installation or execution is required. Users can browse the README for paper titles, links, and categorization.

Highlighted Details

• Over 400 papers are cataloged, covering a wide range of TSF/STF methodologies.
• Includes recent advancements, such as Mamba-based models, LLM applications, and Diffusion Models for TSF.
• Papers are categorized by model architecture (e.g., Transformer, MLP, GNN, SSM) and task type (univariate, multivariate, spatio-temporal, irregular time series).
• Many entries link directly to official code repositories, facilitating reproducibility and practical exploration.

Maintenance & Community

The repository is actively maintained, with recent updates in late 2024. Users are encouraged to contribute via pull requests or issues to suggest new papers or corrections. Links to related "awesome" repositories are provided for further exploration.

Licensing & Compatibility

The repository itself is licensed under the MIT License, allowing for broad use and distribution. The content consists of links to research papers, whose licenses are independent and governed by their respective publishers or authors.

Limitations & Caveats

The primary limitation is that this is a reading list, not an executable framework. While code links are provided, the repository itself does not offer a unified API or benchmarking environment. The categorization is based on the paper's self-description, and some models might fit into multiple categories.