Using ranger R package, while obtaining importance using 'impurity_corrected' option, I get some importance values to be -ve.
My understanding was that the importance values would be from 0 to 1, with 0 meaning not important features (i.e. not adding any value)
Since I get -ve importance values, what do the -ve values mean ? Negatively correlated ? Or worse than 0 (I don't know what that means)
Any reference papers / docs would be helpful
I did check : https://github.com/imbs-hl/ranger/blob/master/R/ranger.R and https://github.com/cran/ranger/blob/master/src/rangerCpp.cpp but could not really arrive at a conclusion.
The reference paper for this method : https://academic.oup.com/bioinformatics/article/34/21/3711/4994791 does not specify what negative values mean
