search

ppdebreuck / modnet

MODNet: a framework for machine learning materials properties
MIT License
81 stars 34 forks source link

Consistent handling of NaN/missing values #28

Closed ml-evs closed 3 years ago

ml-evs commented 3 years ago

We should check this throughout the whole featurization process. Do we consistently replace the nan with some value? here it is zero, but in cleaning it is -1.

_Originally posted by @ppdebreuck in https://github.com/ppdebreuck/modnet/pull/23#discussion_r572877921_

As suggested, we should probably move all this to clean_df, or handle it with a scikit transformer/scaler.

ml-evs commented 3 years ago

Closed by #44