org.neuroph.eval

Class KFoldCrossValidation

java.lang.Object

org.neuroph.eval.KFoldCrossValidation

public class KFoldCrossValidation
extends Object

This class implements multi-threaded cross validation procedure. Splits data set into k subsets (folds), trains the network with data from k-1 and tests with one subset Repeats the procedure k times each time using different subset for testing. TODO: check classifier evaluation, strange results. What about true negative for multi class? Return mean and std for metrics Which network to use as a final result? Best one? Evaluator for classifiers and regressors

Author:

Boris Fulurija, Lukic Sasa Multithreading, Zoran Sevarac, Igor Jovic, Nevena Milenkovic

Constructor Summary

Constructors

Constructor and Description
KFoldCrossValidation(NeuralNetwork neuralNetwork, DataSet dataSet, int numFolds)

Method Summary

Modifier and Type	Method and Description
static ClassificationMetrics.Stats	averageClassificationMetrics(List<ClassificationMetrics.Stats> metricsList)
List<FoldResult>	getResultsByFolds()
void	printFoldResults(ConfusionMatrix confusionMatrix, int foldIdx)
void	printResults(DataSet dataset, ClassificationMetrics.Stats nst, int numfolds)
void	printStats(ConfusionMatrix confusionMatrix)
EvaluationResult	run()
ConfusionMatrix	sumConfusionMatrix(List<ConfusionMatrix> cmList, DataSet dataSet)

Methods inherited from class java.lang.Object

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

Constructor Detail

KFoldCrossValidation

public KFoldCrossValidation(NeuralNetwork neuralNetwork,
                            DataSet dataSet,
                            int numFolds)

Method Detail

run

public EvaluationResult run()
                     throws InterruptedException,
                            ExecutionException

Throws:

InterruptedException

ExecutionException

sumConfusionMatrix

public ConfusionMatrix sumConfusionMatrix(List<ConfusionMatrix> cmList,
                                          DataSet dataSet)

averageClassificationMetrics

public static ClassificationMetrics.Stats averageClassificationMetrics(List<ClassificationMetrics.Stats> metricsList)

printResults

public void printResults(DataSet dataset,
                         ClassificationMetrics.Stats nst,
                         int numfolds)

printFoldResults

public void printFoldResults(ConfusionMatrix confusionMatrix,
                             int foldIdx)

printStats

public void printStats(ConfusionMatrix confusionMatrix)

getResultsByFolds

public List<FoldResult> getResultsByFolds()