When using Pandas DataFrames about boruta_py HOT 3 CLOSED

Yepp atm boruta expects a numpy array for X, but this is made explicit in the docstring of fit():
X : array-like, shape = [n_samples, n_features]
The training input samples.

If you feel this is an important issue, please add this to the fit and I'll review your changes.

Oh you did, that's wonderful, cheers!

from boruta_py.

The examples show pandas going in. I suppose it would be as easy to just update the user doc to show them to only send numpy. I built a 'pandas check' but that has the unfortunate side effect of adding a dependency. It appears that's how sklearn handles it as well though. Toss up, I'll leave you to decide which you like better :)

from boruta_py.

Hi Mike,

Yepp, I wanted it to have a scikit learn interface, so kinda instinctively stuck with the numpy input as sklearn does.. I added a warning to the examples as you recommended, and renamed boruta_py2 to boruta_py_plus.. Also left in your sanity check for pandas dataframes jsut in case. Pandas is pretty common now, it's not a major dependency issue imo..

Thanks again for your valuable input, really appreciate it!

cheers,
Dan

from boruta_py.