« first day (3755 days earlier)   

1:05 AM
howdy all. ISTR there was a question, either here or on datascience.SE, about what variable importance measures actually mean. The answer I saw was that they mean different things depending on what the underlying model is, eg what you get out of an elastic net regression is different to what you get from a boosting model. Can anyone point me to this? Doing a search for variable importance results in a ton of hits, most of which are very generic
 

« first day (3755 days earlier)