I have a multi-class classification dataset that has readings from several places, and multi-class output.
To get understanding of model performance, i am utilizing sklearn's cross-validate method.
I am also utilizing leave-one-group out as a cv technique.
for one of groups, I have one class absent. i.e, for the full dataset (pseudo-code)
y.unique()
[0,1,2,3]
and for one of the splits
y_split.unique()
[0,2,3]
when XGBClassifier tries to fit on this fold, it throws an error:
Invalid classes inferred from unique values of `y`. Expected: [0 1 2], got [0 2 3]
Do you have any hints how I can overcome such behavior?
Apparently, I can not do much with the data - it provided as-is.
