Codecov Report

❌ Patch coverage is 90.90909% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 91.34%. Comparing base (7ff0ca9) to head (9727135).
⚠️ Report is 4 commits behind head on main.

@@            Coverage Diff             @@
##             main      #57      +/-   ##
==========================================
+ Coverage   86.51%   91.34%   +4.82%     
==========================================
  Files           6        6              
  Lines         786      797      +11     
==========================================
+ Hits          680      728      +48     
+ Misses        106       69      -37     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

You should fix the docstring and try to avoid that blatant local overfitting there :). Also, needs to be added to API reference.

yeah I was looking at exactly that right now..😅
getting the actual values on local rn.

hey @martinfleis so i tested it out on local - it highly overfits on include_focal = true , and gives R^2 values as ~0.9999999 - but the values on include_focal = false it performs poorly on the training set and shows some bad, but honest numbers.

And also, for the example on Random Forest Regressor - the values are also overfit and all of them give ~0.8 since it includes_focal = true by default

For now, I've added the actual values in this commit and testing if CI works. I'll change immediately if you say so.

We don't have to show local r2 but can show pred_ instead. In any case, none of the ensemble models should include focal by default. I did set it to False for ensemble classifiers but forgot in RF regressor. Can you update that as well?

@martinfleis shall i make the pred_ changes inside the ensemble.ipynb file too and run again?

No need to change anything apart from that sentence about OOB and "fails" I mentioned above. You can execute it again, yes but it gets executed automatically when building the docs anyway.

@martinfleis yes, changed.