OpenTS:

Democratizing AI for Time series and Spatio-temporal data (AI4TS)

OpenTS is an open, comprehensive AI suite designed to make cutting-edge AI4TS research accessible to all. Built on the AGREE principles—Automation, Generalization, Robustness, Explainability, and Efficiency—it empowers researchers and practitioners with:

  • Benchmarks: Standardized evaluations for fair comparisons and reproducible AI4TS research.
  • Models: Cutting-edge AI4TS models—including Foundation Models, AutoML, and Deep Learning—ready for both academic research and real-world deployment, covering the full-cycle workflow from data governance, analytics, to decision making.
  • Applications: End-to-end solutions across domains such as AIOps, Intelligent Transportation, Virtual Power Plants, AI4DB, FinTech, and Smart Emergency Response.

Developed by the Decision Intelligence Lab at East China Normal University, in collaboration with the Department of Computer Science at Aalborg University, OpenTS advances data-driven decision intelligence by transforming complex data into AI-powered decisions, bridging cutting-edge research with real-world impact.

Benchmarks
(OpenTS-BM)

Fair & comprehensive benchmarking

Foudation Models
(OpenTS-FM)

Time Series Foundation Models (TSFMs) with strong Zero-shot/Few-shot abilities across different tasks

AutoML
(OpenTS-Auto)

Automated model selection, automated ensemble, and neural architecture search

Deep Learning
(OpenTS-DL)

Addressing specific challenges with the AGREE properties

Timeline of AI4TS developed by the DI@ECNU and DESS@AAU

Time series

Spatio-temporal data

Visitors