[Python] How can I use sklearn's make_column_selector to select all valid datetime columns?

I want to select columns based on their datetime data types. My DataFrame has for example columns with types np.dtype('datetime64[ns]'), np.datetime64 and 'datetime64[ns, UTC]'.

Is there a generic way to select all columns with a datetime datatype?


For instance, this works:

from sklearn.compose import make_column_selector
selector = make_column_selector(dtype_include=(np.dtype('datetime64[ns]'),np.datetime64))
selected_columns = selector(df)


But this doesn't (datatype from a pandas df with 'UTC'):

from sklearn.compose import make_column_selector
selector = make_column_selector(dtype_include=np.dtype('datetime64[ns, UTC]')
selected_columns = selector(df)


Compared to numeric data types where you can simply use np.number instead of np.int64 etc.

API-Reference to make_column_selector: LINK

Continue reading...