Class Confusion was was designed to help extrapolate your models decisions through visuals such as graphs or confusion matrices that go more in-depth than the standard plot_confusion_matrix. Class Confusion can be used with both Tabular and Image classification models
To utilize this function, input in the ClassificationInterpretation object as well as a list of classes you want to examine:
from ClassLosses import *
ClassLosses(interp, classList)You can also pass in direct class combinations you want to see as well, just make them a list of tuples as such:
comboList = [('<50k', '>=50k')]
ClassLosses(interp, comboList, is_ordered=True)Please read the Documentation for a guide to how to utilize this function.
Some example outputs: