sktime

sktime is a unified framework for machine learning with time series, built to do for temporal data what scikit-learn did for tabular data. It provides one consistent API for forecasting, time series classification, regression, annotation and anomaly detection, wrapping hundreds of algorithms and dozens of external libraries. It has about 10,100 stars on GitHub and is actively developed in 2026.

Overview

Time series tooling used to mean a different interface for every library — statsmodels here, a bespoke forecaster there. sktime standardizes it: estimators expose the same fit/predict interface, work in pipelines with transformers and tuning tools, and interoperate so you can benchmark a classical ARIMA-style model against deep learning forecasters and gradient-boosted regressors in a few lines. The framework interfaces with libraries such as statsforecast, Prophet-style models, tslearn and many neural architectures, and adds time-series-specific machinery like temporal cross-validation splitters, forecasting horizons and reduction strategies that turn forecasting into supervised learning. It is a core piece of the open-source time series ecosystem, maintained by a broad community under the sktime organization. sktime also sits at the center of a wider ecosystem: its interfaces are reused by other forecasting and classification libraries that want scikit-learn compatibility without reimplementing it, and its estimator lookup provides a searchable registry of hundreds of algorithms with their capabilities and tags. The project runs a structured open-source programme that onboards new contributors every year, keeping maintenance broad rather than dependent on a few individuals. Documentation includes tutorials, extension templates for adding your own estimators, and migration guides between releases.

Key Features

• Unified scikit-learn-style API for all time series tasks
• Forecasting, classification, regression and anomaly detection
• Hundreds of estimators plus interfaces to major external libraries
• Pipelines, ensembling and hyperparameter tuning support
• Time-aware cross-validation and forecasting horizons
• Extensible estimator registry with conformance testing

💳 Pricing

sktime is completely free and open source under the BSD 3-Clause license, for academic and commercial use alike.

Who Is It For

sktime is for data scientists and ML engineers working with temporal data — demand forecasting, sensor analytics, finance, healthcare signals — who want one coherent toolbox instead of a pile of incompatible libraries.