Try out categorical mode with xgboost

[marco-c]

See https://xgboost.readthedocs.io/en/stable/tutorials/categorical.html.

[gothwalritu]

Hello marco-c! I am interested in this issue. Could you please share more details on this?

[marco-c]

@gothwalritu it might be nice to continue with #3717 since it looks like you're close to completion.

If you want to explore this too, you could pick one of the models already implemented in bugbug/models, and try to use the https://xgboost.readthedocs.io/en/stable/tutorials/categorical.html.
You'd need to convert the categorical features to actually be categorical by doing .astype("category"), and then enable the functionality by passing enable_categorical=True to the XGBoost classifier constructor.

You can convert features to categorical in

bugbug/bugbug/bug_features.py

Line 601 in c598ee7

class BugExtractor(BaseEstimator, TransformerMixin):

. We need to find a nice way to do it, but for testing you could just do it column by column on the dataframe here

bugbug/bugbug/bug_features.py

Line 689 in c598ee7

return pd.DataFrame(apply_transform(bug) for bug in bugs_iter)

.

[gothwalritu]

Yes, I will close #3717 first. I just wanted to get myself acquainted with this one in the meantime. Thanks for your reply. BTW Happy New Year!!

[marco-c]

@gothwalritu happy new year to you too!

[suhaibmujahid]

Duplicate of #2819