Find discriminative features based on annotation
There is something that I need for almost every project: Given some class annotations I would like to know which of the measured features are most useful to distinguish the different classes. Could we add something for this to the IDE?
For example, maybe a visualisation (could be pure text if there is nothing graphical) of the best decision tree? This would have the additional benefit that one could also see at which threshold values the features should be used to distinguish the classes.
CellProfiler Analyst does something very similar: https://www.ncbi.nlm.nih.gov/books/NBK126174/figure/hcsimage.F8/
Edited by Christian Tischer