sktime - python toolbox for time series
summary
sktime is scikit-learn compatible and claims to be easily extensible and interoperable with the PyData stack. the tutorial shows runnable notebooks (including a binder link) and demos concrete features added since 2022: sktime is scikit-learn compatible, graphical pipelines (experimental), improved parallelization with backend options and broadcasting across multivariate/hierarchical panels, pipelines that handle transform + inverse, probabilistic predictions that return distribution objects, estimator search, blueprint serialization and MLflow export — these features are built and demonstrated, with graphical pipelines marked experimental.
AI-written summary · about
chapters (8)
- 0:00Introduction to sktime
- 11:38Basic forecasting
- 27:31Exogenous features for forecasts
- 47:44Multivariate and hierarchical forecasting
- 1:00:31Graphical pipelines
- 1:14:53Advanced modeling features
- 1:23:22Deployment capabilities
- 1:27:40Summary
related talks
30msktime: python toolbox for time series- next-generation AI
01h 26mSktime—Python Toolbox for Time Series: How to Implement Your Own Estimator
26mTime Series Analysis in Python: Easy Tools for Scientific Insight
01h 27mIntroduction to Machine Learning with Time Series
32mAdopting Open-Source Tools for Time Series Forecasting
01h 04ma workbench for creating scikit-learn like parametric objects and libraries
01h 29msktime: time series anomaly detection, changepoint detection, segmentation
34mTime series for scikit-learn people