A list of train objects. If the model fails to build,
it is dropped from the list.
Arguments
...
arguments to pass to train. Don't use the formula interface, its slower
and buggier compared to the X, y interface. Use a data.table for X.
Particularly if you have a large dataset and/or many models, using a data.table will
avoid unnecessary copies of your data and can save a lot of time and RAM.
These arguments will determine which train method gets dispatched.
trControl
a trainControl object. If NULL, will use defaultControl.
methodList
optional, a character vector of caret models to ensemble.
One of methodList or tuneList must be specified.
tuneList
optional, a NAMED list of caretModelSpec objects.
This much more flexible than methodList and allows the
specification of model-specific parameters (e.g. passing trace=FALSE to nnet)
metric
a string, the metric to optimize for. If NULL, we will choose a good one.
continue_on_fail
logical, should a valid caretList be returned that
excludes models that fail, default is FALSE
trim
logical should the train models be trimmed to save memory and speed up stacking