I have been searching around for a Scikit-learn package which calculates the specificity for multiclass problems but I can't seem to find one.
I had a looked at sklearn.metrics.recall_score but this package does not calculate the specificity for multiclass problems, in particular setting pos_label=0 does not work for the multiclass case.
sklearn.metrics.precision_recall_fscore_support and sklearn.metrics.classification_report also doesn't appear to support the calculation of specificity for multiclass problems.
If Scikit-learn intentionally did not include the specificity for multiclass problems in their packages, is there a reason in doing so?
