org.tribuo.evaluation

Class KFoldSplitter<T extends Output<T>>

java.lang.Object

org.tribuo.evaluation.KFoldSplitter<T>

Type Parameters:

T - the type of the examples that make up the dataset to be split

public class KFoldSplitter<T extends Output<T>>
extends Object

A k-fold splitter to be used in cross-validation.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	KFoldSplitter.TrainTestFold<T extends Output<T>> Stores a train/test split for a dataset.

Field Summary

Fields

Modifier and Type	Field and Description
protected int	nsplits
protected SplittableRandom	rng
protected long	seed

Constructor Summary

Constructors

Constructor and Description
KFoldSplitter(int nsplits) Builds a k-fold splitter using Trainer.DEFAULT_SEED as the seed.
KFoldSplitter(int nsplits, long randomSeed) Builds a k-fold splitter.

Method Summary

Modifier and Type	Method and Description
Iterator<KFoldSplitter.TrainTestFold<T>>	split(Dataset<T> dataset, boolean shuffle) Splits a dataset into k consecutive folds; for each fold, the remaining k-1 folds form the training set.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

nsplits

protected final int nsplits

seed

protected final long seed

rng

protected final SplittableRandom rng

Constructor Detail

KFoldSplitter

public KFoldSplitter(int nsplits,
                     long randomSeed)

Builds a k-fold splitter.

Parameters:

nsplits - The number of folds.

randomSeed - The RNG seed.

KFoldSplitter

public KFoldSplitter(int nsplits)

Builds a k-fold splitter using Trainer.DEFAULT_SEED as the seed.

Parameters:

nsplits - The number of folds.

Method Detail

split

public Iterator<KFoldSplitter.TrainTestFold<T>> split(Dataset<T> dataset,
                                                      boolean shuffle)

Splits a dataset into k consecutive folds; for each fold, the remaining k-1 folds form the training set.

Note: the first nsamples % nsplits folds have size nsamples/nsplits + 1 and the remaining have size nsamples/nsplits, where nsamples = dataset.size().

Parameters:

dataset - The full dataset

shuffle - Whether or not shuffle the dataset before generating folds

Returns:

An iterator over TrainTestFolds