Implements sample_weight and optional permutation and SHAP importance, categorical features, boxplot

[ThomasBury]

Hi,

It took me a while but finally found the time to work on the continuation of the discussion https://github.com/scikit-learn-contrib/boruta_py/pull/77

Meaning:

• Not introducing new dependencies, a check import is performed if the User wants to use SHAP or get the matplotlib boxplot
• Permutation importance (sklearn) is also implemented but optional and easy to switch off
• Categorical features are encoded if any (optional)
• sample_weight can now be passed to the fit method
• A notebook illustrates the changes and compares the original Boruta_py and the new features
• Add a note in the readme

[danielhomola]

Hi Thomas,

Thanks for this! Will try to find time in the next few weeks to go through it (it's quite a lot). To start with however, can we make sure that no .idea and .ipython-checkpoints file are committed? Thanks!

[ThomasBury]

Hi Thomas,

Thanks for this! Will try to find time in the next few weeks to go through it (it's quite a lot). To start with however, can we make sure that no .idea and .ipython-checkpoints file are committed? Thanks!

Sorry, they were remains of the previous ignore, .idea and checkpoints are removed. I hope the notebook will be helpful, do not hesitate if you have any questions/remarks.

Thanks

[erikvdp]

This seems like a really cool PR! Is there any chance that it will get merged soon?

[ThomasBury]

This seems like a really cool PR! Is there any chance that it will get merged soon?

Thanks @erikvdp, meanwhile, you might have a look at https://github.com/ThomasBury/arfs implementing those and more (although I still think it'd best to integrate the changes related to boruta in the official boruta_py ^^)

[MauritsDescamps]

Any chance this will be merged? Would really like to try out Boruta with Shap feature importance

[ThomasBury]

Any chance this will be merged? Would really like to try out Boruta with Shap feature importance

Hi @MauritsDescamps, I built the ARFS package to provide those features for Boruta (and much more). In the ARFS pkg, you'll find 3 different methods for performing all relevant feature selection. I called the evolution of Boruta: "Leshy" and it provides the features of this PR. There are notebooks that explain step by step how to use it and what are the differences.

you can test it by simply pip install -U arfs, there is a brand new release (version 1.0.2)