I'm working with a high dimensional data set including 80 features, for which I try to select the most important ones. In order to do so, I apply the random forest method from R package 'ranger' to the data and the provided function 'variable.importance'. It returns the features attached with a score of signed values absolutly lower than approximatly 1/10. Does anybody knows, how these values be interpreted and utilize correctly for proper feature importance selection? Unfortunatly the documentation of the R package is not too detailed.
Thank you in advance!
I already applied the package 'ranger' and read the documentation.