de.jungblut.classification.eval

Class EvaluationSplit

java.lang.Object

de.jungblut.classification.eval.EvaluationSplit

public class EvaluationSplit
extends Object

Split data class that contains the division of train/test vectors. If no division is yet there, you can use the create(DoubleVector[], DoubleVector[], float, boolean) method.

Author:

thomas.jungblut

Constructor Summary

Constructors

Constructor and Description
EvaluationSplit(de.jungblut.math.DoubleVector[] trainFeatures, de.jungblut.math.DoubleVector[] trainOutcome, de.jungblut.math.DoubleVector[] testFeatures, de.jungblut.math.DoubleVector[] testOutcome) Sets a split internally.

Method Summary

Modifier and Type	Method and Description
static EvaluationSplit	create(de.jungblut.math.DoubleVector[] features, de.jungblut.math.DoubleVector[] outcome, float splitFraction, boolean random) Creates a new evaluation split.
static EvaluationSplit	createStratified(de.jungblut.math.DoubleVector[] features, de.jungblut.math.DoubleVector[] outcome, float splitFraction, boolean random) Creates a new stratified evaluation split.
de.jungblut.math.DoubleVector[]	getTestFeatures()
de.jungblut.math.DoubleVector[]	getTestOutcome()
de.jungblut.math.DoubleVector[]	getTrainFeatures()
de.jungblut.math.DoubleVector[]	getTrainOutcome()

Methods inherited from class java.lang.Object

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

Constructor Detail

EvaluationSplit

public EvaluationSplit(de.jungblut.math.DoubleVector[] trainFeatures,
                       de.jungblut.math.DoubleVector[] trainOutcome,
                       de.jungblut.math.DoubleVector[] testFeatures,
                       de.jungblut.math.DoubleVector[] testOutcome)

Sets a split internally.

Method Detail

getTrainFeatures

public de.jungblut.math.DoubleVector[] getTrainFeatures()

getTrainOutcome

public de.jungblut.math.DoubleVector[] getTrainOutcome()

getTestFeatures

public de.jungblut.math.DoubleVector[] getTestFeatures()

getTestOutcome

public de.jungblut.math.DoubleVector[] getTestOutcome()

create

public static EvaluationSplit create(de.jungblut.math.DoubleVector[] features,
                                     de.jungblut.math.DoubleVector[] outcome,
                                     float splitFraction,
                                     boolean random)

Creates a new evaluation split.

Parameters:

features - the features of your classifier.

outcome - the target variables of the classifier.

splitFraction - a value between 0f and 1f that sets the size of the trainingset. With 1k items, a splitPercentage of 0.9f will result in 900 items to train and 100 to evaluate.

random - true if data needs shuffling before.

Returns:

a new EvaluationSplit.

createStratified

public static EvaluationSplit createStratified(de.jungblut.math.DoubleVector[] features,
                                               de.jungblut.math.DoubleVector[] outcome,
                                               float splitFraction,
                                               boolean random)

Creates a new stratified evaluation split. Sampling is done based on the max index of the outcome classes (assumes one-hot encoding, or zero/one encoding for binary classes). This class does not keep the relation of the original outcome vectors to their features, thus every mutual information stored for both should be included in the feature vector.

Parameters:

features - the features of your classifier.

outcome - the target variables of the classifier.

splitFraction - a value between 0f and 1f that sets the size of the trainingset. With 1k items, a splitPercentage of 0.9f will result in 900 items to train and 100 to evaluate.

random - true if data needs shuffling before.

Returns:

a new EvaluationSplit.