Hi,
In the third parameter of a preprocessor in ColumnTransformer, you use either the columns names categorical_columns or the columns selector categorical_columns_selector.
Does it make any difference ?
How the ColumnTransformer is made robust in relation with this parameter ?
It is the same. At the end the ColumnTransformer will be applied to data, which is equivalent to internally defining
categorical_columns = categorical_columns_selector(data).
Maybe the only difference is that making the definition externally allows you to inspect the names of the columns that are being selected.
Ok, thank you for your answer !