sktime provides an easy-to-use, flexible and modular open-source framework for a wide range of time series machine learning tasks. It offers scikit-learn compatible interfaces and model composition tools, with the goal to make the ecosystem more usable and interoperable as a whole. We build and sustain an open, diverse and self-governing community, welcoming new contributors from academia and industry through instructive documentation, mentoring and workshops.


Our Roadmap describes how we try to achieve our mission.

Motivation: Why we develop sktime#

There are a number of reasons why we develop sktime, here are the most important ones:

  • Easier development of ML :term:`applications <application>` with time series, by making the ecosystem more interoperable and usable as a whole, and by providing an easy-to-use and readable way of specifying and applying algorithms.

  • Easier ML algorithm development, from research to implementation and testing.

  • Better ML algorithm research, by facilitating reproducibility and fair evaluation and comparison of different algorithms.

  • Easier teaching and learning for ML with time series.

  • Clearer time series methodology, especially more consistent definitions, terminology and notation across learning tasks.

  • More collaboration and innovation, by bringing together developers, practitioners and domain experts in a single project, by exploring a new community-driven model for the data science innovation cycle from methodology research to software development and deployment, and by offering alternative career paths for junior researchers and practitioners.