support for grouped DataFrames in featurizers about matminer HOT 5 CLOSED

(maybe @WardLT @JFChen3 for the latter improvements to featurize_dataframe() ?)

from matminer.

Nice! I didn't realize you could group columns this way. The code looks decently straightforward. I could give it a shot for marking which featurizer a feature comes from for MultiFeaturizer.

from matminer.

@ardunn do you consider this complete? if so can you close?

from matminer.

@computron This is definitely not complete. I removed it from a previous PR because it made the PR too complicated. It is on the "todos" though

Edit:

@WardLT if you have a good idea/gameplan for implementing this, feel free to take it over. I am finding that for basic multiindexing it is okay, but when using inplace or with MultiFeaturizer, multiindexing introduces a lot of code complexity for such a small increase in utility.

from matminer.

How about skipping MultiFeaturizer and inplace for now?

• For MultiFeaturizer give it all the same group heading, "multifeaturizer" or something
• For inplace, just leave the same group that you had before and don't mess with it

Would that allow the issue to at least be partially solved and usable?

from matminer.