Improve reporting of receiving operating curves (ROC) and precision-recall curves (PRC). Each runs result should be save to a dataframe for inspection and calculation of statistics. This then also enables ROC and PRC curves with confidence bands.
- Check how this was done in ALD study
- Provide bootstrapped confidence interval as done by seaborn?
Make is possible to specify desired feature set, start at njab.sklearn.find_n_best_features