9 Comments
User's avatar
Joe Corliss's avatar

There is a chance (however small) that the random feature will have moderate or high importance and cause us to drop useful features. I would want to run the process 100+ times and drop the features that are identified the most often. But this would require many model fits. Overall, I would prefer to calculate the permutation importances once and see if there's a clear threshold for features worth keeping.

Expand full comment
Kumar Abhishek's avatar

Thank you for this post. How do you draw these nice animated arrows? Is there a library for exaclidraw or do you use any other tool?

Expand full comment
Avi Chawla's avatar

I used draw.io, Abhishek.

Expand full comment
morgul's avatar

This is like very bare-bones, and worse, version of Boruta's algorithm

Expand full comment
Kyle Gilde's avatar

Is there a scikit learn compatible package that will do this for us?

Expand full comment
Avi Chawla's avatar

I don't think so. But all you need to do is add a new feature and run the ML algorithm. So it's no that much of an effort :)

I will share if I find any plug-and-play library for this.

Expand full comment
George Carter's avatar

Great article. Any recommendations for creating the "random feature" to assure the Probe Method gets the best result?

Expand full comment