Closed
Description
Describe the workflow you want to enable
Inspired by #26391 I think that support for missing values for ExtraTrees regressor and classifier should/could also be provided.
Describe your proposed solution
I think a foundational work is already provided by @thomasjpfan in #26391 and besides tests and documentation to enable nan handling it is enough to modify sklearn/tree/_classes.py:
For ExtraTreeRegressor add method:
def _more_tags(self):
# XXX: nan is only support for dense arrays, but we set this for common test to
# pass, specifically: check_estimators_nan_inf
allow_nan = self.criterion in {
"squared_error",
"friedman_mse",
"poisson",
}
return {"allow_nan": allow_nan}
For ExtraTreeClassifier add method:
def _more_tags(self):
# XXX: nan is only support for dense arrays, but we set this for common test to
# pass, specifically: check_estimators_nan_inf
allow_nan = self.criterion in {
"gini",
"log_loss",
"entropy",
}
return {"multilabel": True, "allow_nan": allow_nan}
I've run the code locally, and it appears to be functioning as expected. However, I must emphasize that my testing was not exhaustive, and I might have overlooked some obvious aspects.
Describe alternatives you've considered, if relevant
No response
Additional context
No response
Metadata
Metadata
Assignees
Type
Projects
Status
Done