This page contains guides and links to resources that show how OpenSTEF can be used.
Pipelines - high level functionality#
Pipelines (see concepts for definition) offer an easy way to use OpenSTEF for training models, generating forecasts, and evaluation of forecasting performance.
The following pipelines are available:
The easiest way to get started and get familiar with pipelines is to have a look at this GitHub repository that contains an assortment of examples Jupyter notebooks, including example data. Each of these example notebooks can be ran locally without any setup required, apart from the installation of the OpenSTEF package.
Usage of the train model pipeline is demonstrated in this example Jupyter Notebook.
Usage of the create forecast pipeline is demonstrated in this example Jupyter Notebook.
Usage of the train model and forecast backtest pipeline is demonstrated in multiple notebooks, for instance this example Jupyter Notebook.
The notebooks mentioned above are aimed towards a backtesting setting. More in depth information on how to use and implement the pipelines in an operational setting, including code examples, is provided on the following page:
Forecasting application - full implementation#
For those who wish to set up a fully functioning forecasting application that is ready to be used in an operational setting, a GitHub repository with a reference implementation is available. The example implementation includes databases, a user interface, and example data. More information on what the architecture of such an application could look like can be found here.
Screenshot of the operational dashboard showing the key functionality of OpenSTEF. Dashboard documentation can be found here.
Example Jupyter notebooks#
Jupyter Notebooks demonstrating some of OpenSTEF’s main functionality can be found at: OpenSTEF/openstef-offline-example.