We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Franz Kiraly: sktime - python toolbox for time series: advanced forecasting - probabilistic, glob...
Develop advanced time series forecasting techniques with SKTime, a Python toolbox, and explore its hierarchical and probabilistic methods, including compositors, evaluation, and metrics, perfect for panel data and global forecasting.
- SKTime is a Python toolbox for time series forecasting, aiming to provide a vast range of advanced forecasting techniques, including probabilistic and hierarchical methods.
- It is designed to be a Scikit-learn-like toolbox, with similar interfaces and architecture for learning with time series data.
- Hierarchical forecasting allows modeling multiple levels of the hierarchy, such as product-level and total-level forecasts, and can be used for coping with irregular sampling or missing data.
- Probabilistic forecasting can be achieved through various methods, including compositors, such as pipelines and transformers, which allow combining multiple forecasts and integrity functions.
- SKTime provides an evaluation interface for evaluating both point and interval forecasts, as well as quantile forecasts.
- The toolkit also offers a range of metrics for interval and quantile forecasting, including the pinball loss.
- It supports various formats, including panel data, which allows for multiple observations over time of multiple systems or independent experimental units.
- SKTime is a community-based project, and contributors are encouraged to participate in community operations and events.
- Hierarchical global forecasting and panel forecasting are some of the key potential applications of SKTime.