fixing column masking & typo

[rcarneva]

Two find-and-replaces:

1. Column masking in the exported pipeline wasn't working correctly for me with SelectKBest on a model since the mask was coming through as a numeric but some of the columns were named. Updated [mask] -> .iloc[:, mask] and it seems to work now, though more testing would be good if it was working as expected on your end.
2. indeces -> indices

[rhiever]

Thanks @rcarneva! Both versions work fine on my end, though I believe the .iloc[:, mask]syntax is "more correct." What version of Python/pandas are you running?

I'll merge this in the meantime.

[rcarneva]

Python 2.7.10 / pandas 0.17.1

[rhiever]

Interesting. I'm on Python 3.5 / pandas 0.17.1 and both cases work.

[rcarneva]

Think I tracked it down. The two features in my training set had (integer) names 1 and 2, and so pandas seemed to have trouble applying the integer mask. Renaming the columns to strings made it work with the original code, but the updated version with iloc seems to work in both situations, so is probably a safer bet.

BTW, really neat idea and project.

[rhiever]

I had a feeling that was it. Thanks for looking into it!

BTW, really neat idea and project.

Thanks! Happy to have you involved more if you have any ideas or any of the open issues catch your fancy.