workflow: A workflow to Evaluate the Performance of a List of Methods

In SkadiEye/ITRlearn: Individualized Treatment Rule Learning

Description

In this procedure, if a validation set is provided, all methods in the list will be implemented once and the performance is evaluated using the validation set.

Usage

workflow(trainObj, validObj = NULL, k.fold = 10, n.iter = 10, cvCtrl.list,
  unbalance.trt = c("None", "over", "under")[1], cutoff = 0.5)

Arguments

trainObj	A RegObj or wClsObj containing the training set.
validObj	A RegObj or wClsObj containing the validation set.
k.fold	The number of folds in the cross validation.
n.iter	The number of interations of k-fold cross validations.
cvCtrl.list	A list of lists of arguments that specifies a regression or classification method. Each list of arguments is handled by itrCtrlPanel.
unbalance.trt	A character string that specifies if sampleing techniques should be applied to data with unbalanced numbers for each class.
cutoff	Cutoff used when predicting the current fold that not used in training the model, passed to cutoff.calc function.

Details

If the validation set is missing, an n iterations of k-fold cross validation will be performed to evaluate each method.

The final performance scores will be reported. And all final models will be saved.

Value

A workflow-class object.

See Also

cutoff.calc
reg
wcls
cvGrid
ensemble
cvKfold

SkadiEye/ITRlearn documentation built on May 24, 2019, 1:31 a.m.