caret's helper functions for feature selection: caretSBF and caretFuncts

Question

caret provides functions for nested feature selection, via caretSBF and caretFuncs.

caretSBF: for Selection By Filter

lmSBF
rfSBF
treebagSBF
ldaSBF
nbSBF

caretFuncs: for Recursive Feature Selection -- (I assume it's just for historical reasons that the caretFuncs are not called caretRFE?)

lmFuncs
rfFuncs
treebagFuncs
ldaFuncs
nbFuncs
gamFuncs
lrFuncs

Why is this list of helper functions so much more limited than caret's full model list? For example, there are no tailored feature selection methods for glmnet's elastic net method, kernlab's svm method, etc.

Is it because:

• these other models work fine with the generic case functions (i.e. caretSBF and caretFuncs)? - though I've run into trouble trying to get glmnet working with caretSBF...
• these other models are not thought to require 'extra help' in feature selection? - I guess a method like elasticnet is thought to do feature selection well, but other methods like SVM certainly do not...
• the package maintainer just hasn't gotten around to writing the specialized functions for other methods? - In which case, could/should the community contribute these helper functions?

Thanks for reading and thanks to the caret package maintainer(s) for the excellent work.

Answer 1

It is a combination of #2 and #3. It does't make sense for models that implicitly do feature selection (e.g. rpart, glmnet etc). Also, I haven't had time to expand the list (but code submissions are always welcome).

Max