Imputers

[tresink]

Hi,

First of all, congrats on a nice package! I read the medium article and immediately tried it.

One thing I am wondering is, is there a way to use imputers from sklearn with pdpipe? Or could you give an example on how to use any class with the fit, transform and fit_transform method with pdpipe?

Kind regards,

Tim

[shaypal5]

Hey @tresink ,

Thank you for your kind words. :)

Yes, it should be fairly easy to write a small generic adapter around sklearn imputers. Basically the only thing missing is casting the data structure from a numpy array back to a dataframe.

I think one could even write a generic wrapper doing this to np.ndarray-to-np.ndarray sklearn transformer.

I'll try to get to it soon, but if you want to take a jab at this, take a look at the documentation section that talks about creating additional stages:
https://pdpipe.github.io/pdpipe/doc/pdpipe/#creating-additional-stages

Especially at the part about ad-hoc pipeline stages...

Cheers,
Shay

[shaypal5]

@carbonleakage

This is also a great first issue.

It can also lead to writing, as a second stage, a generic column-based stage wrapper for any matrix-to-matrix sklearn transformer. What do you say? :)