Pandas fragmentation warning with AddMissingIndicator

[jose-cano]

Pandas fragmentation warning with AddMissingIndicator.

Tested behavior when using AddMissingIndicator in a pipeline with cross validation.

Assuming the fix would be to change line 164 and 165 of the AddMissingIndicator class from:

indicator_names = [f"{feature}_na" for feature in self.variables_]
X[indicator_names] = X[self.variables_].isna().astype(int)

To something like:

indicator_df = (X[self.variables_].isna().astype(int).rename(columns=lambda c: f"{c}_na"))
X = pd.concat([X, indicator_df], axis=1)