A Python module to provide model validation and forecast verification tools, including a set of convenient plot functions. A selection of capabilites provided by PyForecastTools includes:
- Accuracy and bias metrics for continuous predictands
- Unscaled/absolute measures
- Relative measures
- Scaled measures
- 2x2 and NxN contingency table classes
- Wide range of contingency table metrics and scores
- Multiple methods of calculating confidence intervals on scores
- Convenient plotting for visually comparing models and data
- Quantile-Quantile plots
- Taylor diagrams
- ROC curves
- Reliability diagrams
The module builds on the scientific Python stack (Python, Numpy, MatPlotLib) and uses the dmarray class from SpacePy's datamodel.
SpacePy is available through the Python Package Index, MacPorts, and is under version control at github.com/spacepy/spacepy If SpacePy is not available a reduced functionality implementation of the class is provided with this package.
PyForecastTools is available through the Python Package Index and can be installed simply with
pip install PyForecastTools --user
To install (local user), run
python setup.py install --user
After installation, the module can then be imported (within a Python script or interpreter) by
import verify
For help, please see the docstrings for each function and/or class.
Additional documentation is under development using Github pages at drsteve.github.io/PyForecastTools, and source for this is in the docs folder.