Package com.bayesserver.data

Class CrossValidation

java.lang.Object

com.bayesserver.data.CrossValidation

public final class CrossValidation
extends Object

Allows test metrics/scores to be calculated using cross validation.

Method Summary

Modifier and Type	Method	Description
static double	combine(Iterable<CrossValidationTestResult> testResults, CrossValidationCombineMethod method)	Provides standard ways of combining numeric test results from a number of partitions.
static CrossValidationScore[]	kFold(int partitionCount, int testMetricCount, CrossValidationActions actions)	Performs k-fold cross validation.
static List<CrossValidationOutput>	kFoldList(int partitionCount)	Gets a list of training and test DataPartitioning instances for each partition.

Methods inherited from class java.lang.Object

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

Method Detail

kFoldList

public static List<CrossValidationOutput> kFoldList(int partitionCount)

Gets a list of training and test DataPartitioning instances for each partition. This method is an alternative to the callback based kFold(int, int, com.bayesserver.data.CrossValidationActions) method.

Parameters:

partitionCount - The number of k-fold partitions.

Returns:

Training and test DataPartitioning instances for each partition.

kFold

public static CrossValidationScore[] kFold(int partitionCount,
                                           int testMetricCount,
                                           CrossValidationActions actions)

Performs k-fold cross validation.

Parameters:

partitionCount - The number of partitions to use.

testMetricCount - The number of test metrics being calculated.

actions - User supplied actions for the cross validation process.

Returns:

A score for each test metric.

combine

public static double combine(Iterable<CrossValidationTestResult> testResults,
                             CrossValidationCombineMethod method)

Provides standard ways of combining numeric test results from a number of partitions. These can be used in the combine each phase of k-fold cross validation.

Parameters:

testResults - The test results. Typically one for each test partitioning.

method - The method to be used to combine the test results.

Returns:

The overall cross validation score, which can be used to compare models.