TrainLEs: Train Local Experts

In nnormandin/localexpeRt: Binary Local Expert Regression

Description

This function trains a list of local expert models on a binary matrix.

Usage

TrainLEs(x, bincols, trControl = NULL, method = "lda", n.folds = 5,
  metric = "Kappa", JIT = FALSE, ...)

Arguments

x	The feature matrix
bincols	Binary target matrix created with 'BinCols()' function
trControl	Optional argument to specify a train control object- defaults to cross validated
method	Type of learning algorithm used for induction- defaults to lda
n.folds	Number of folds if default train control object is used- defaults to 5
metric	Optimization metric; can be either "Accuracy" or "Kappa"
JIT	Whether or not just-in-time compilation is enabled
...	Additional parameters to pass to caret::train

Details

This function calls the train function independently on each column of the binary matrix supplied to the object supplied to the bincols argument, which should have been generated by the BinCols function. The default optimization metric is the kappa value, but accuracy is also available. For the method argument, only classification algorithms can be used (as opposed to regression).

Value

Returns a list containing train type objects. List length is dependent on the number of columns in the matrix supplied as the argument to bincols

nnormandin/localexpeRt documentation built on May 23, 2019, 9:29 p.m.