Classification performance: 95% intervals
Issue #676
resolved
In the classification workflow, after selecting “Cross validation with n-folds”, how are the 95% confidence intervals that are reported in the resulting accass.html files, calculated? The doc (at least the one in https://enmap-box.readthedocs.io/en/latest/usr_section/usr_manual/processing_algorithms/accuracy_assessment/cross-validated_regressor_performance.html) does not mention it. Perhaps just including a link to the involved scikit-learn module would be sufficient.
Comments (2)
-
-
- changed status to resolved
- Log in to comment
CI are not calculated via scikit learn.
In the current (stabel) version we implemented the formulars from: https://www.sciencedirect.com/science/article/abs/pii/S0034425714000704
In the (experimental) version released two days ago, we use: https://doi.org/10.1080/01431161.2014.930207.
Note that in the (experimental) release we now have two versions of the Classification Workflow app. The “Classic” Version still uses the first implementation.