createCVSplits {aslib}

Create cross-validation splits for a scenario.

Description

Create a data.frame that defines cross-validation splits for a scenario,

and potentially store it in an ARFF file.

The mlr package is used to generate the splits, see makeResampleDesc and makeResampleInstance.

Usage

createCVSplits(asscenario, reps = 1L, folds = 10L, file = NULL)

Arguments

asscenario

[ASScenario]
Algorithm selection scenario.

reps

[integer]
CV repetitions. Default is 1.

folds

[integer]
CV folds. Default is 10.

file

[character]
If not missing, where to save the returned splits as an ARFF file via write.arff. Default is no saving.

Value

[data.frame]. Splits as defined in the algorithm benchmark repository specification text. Has columns: “instance_id”, “fold”, “rep”. Defines which instances go into the test set for each replication / fold during CV. The training set are the remaining instances, in exactly the order as given by the data.frame for the current repetition.