Entering edit mode
7.9 years ago
Xiaokang
▴
80
Boruta will decide a feature important or not comparing its importance with the “shadow”. So when you have enough runs, all the features will be divided into two groups: important and unimportant.
But I’m more interested in the rank of all the features. With the function attStats(), I can get the meanImp of the Boruta object. Can I rank the features according to their ‘meanImp’? Which means that I consider a feature with high "meanImp” value as more important. Is that reasonable?
Thank you in advance.