* This functions aims to replace the first section of the variable selection and relation functions by creating a single reusable object which contains the random forest `ranger::ranger()` model, as well as the `trees` list with layers and surrogates added.
* Returns a `RFSurrogates` object, which serves as the base object for later analysis.
* Returns a `RandomForestSurrogates` object, which serves as the base object for later analysis.
* Additional `...` params are passed directly to `ranger::ranger`.
*`s.pct` is a helper for calculating the number of surrogates as a fraction of number of variables (Default: 0.01). `s` can be set to overwrite this default.
*`mtry` supports the following values:
...
...
@@ -13,7 +14,7 @@
* A `numeric` value for `mtry`.
* The default is `"^3/4"`.
*`type` also uses `match.arg()` and still defaults to `"regression"`.
*`num.threads`transparently defaults to `parallel::detectCores()`.
*`num.threads`defaults to 1.
*`permutate` will, if set to `TRUE`, apply random permutation to the data in each feature. (This is used in permutation importance approaches.)
*`seed` is now a strongly recommended optional parameter (issuing a warning whenever it is not set).
* Setting `seed` will cause a call to `set.seed()` when permutating. It is also used as the `seed` param of the `ranger::ranger()` call.