moa.classifiers.trees

Class FIMTDD.FIMTDDSplitNode

java.lang.Object

moa.AbstractMOAObject

moa.classifiers.trees.HoeffdingTree.Node

moa.classifiers.trees.HoeffdingTree.SplitNode

moa.classifiers.trees.FIMTDD.FIMTDDSplitNode

All Implemented Interfaces:

Serializable, FIMTDD.AdaptationCompatibleNode, MOAObject

Enclosing class:

FIMTDD

public static class FIMTDD.FIMTDDSplitNode
extends HoeffdingTree.SplitNode
implements FIMTDD.AdaptationCompatibleNode

A modified SplitNode method implementing the extra information regarding it's parent, and the ability to track the error rate and perform Page-Hinckley tests

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected FIMTDD	alternateTree
protected DoubleVector	lossStatistics
protected HoeffdingTree.Node	parent
protected double	PHmT
protected double	PHMT
protected int	previousWeight
protected int	weightSeen

Fields inherited from class moa.classifiers.trees.HoeffdingTree.SplitNode

children, splitTest

Fields inherited from class moa.classifiers.trees.HoeffdingTree.Node

observedClassDistribution

Constructor Summary

Constructors

Constructor and Description
FIMTDD.FIMTDDSplitNode(InstanceConditionalTest splitTest, double[] classObservations) Create a new FIMTDDSplitNode

Method Summary

Methods

Modifier and Type	Method and Description
HoeffdingTree.Node	getParent() Return the parent node
void	learnFromInstance(weka.core.Instance inst, FIMTDD ht, boolean growingAltTree) This method has been added to the SplitNode subclass to allow alternate trees to be grown and compared to the current tree, to allow adaptation in the face of concept drift
boolean	PageHinckleyTest(double error, double threshold) Check to see if the tree needs updating
void	setParent(HoeffdingTree.Node parent) Set the parent node

Methods inherited from class moa.classifiers.trees.HoeffdingTree.SplitNode

calcByteSize, calcByteSizeIncludingSubtree, describeSubtree, filterInstanceToLeaf, getChild, instanceChildIndex, isLeaf, numChildren, setChild, subtreeDepth

Methods inherited from class moa.classifiers.trees.HoeffdingTree.Node

calculatePromise, getClassVotes, getDescription, getObservedClassDistribution, observedClassDistributionIsPure

Methods inherited from class moa.AbstractMOAObject

copy, copy, measureByteSize, measureByteSize, toString

Methods inherited from class java.lang.Object

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

Field Detail

parent

protected HoeffdingTree.Node parent

PHmT

protected double PHmT

PHMT

protected double PHMT

alternateTree

protected FIMTDD alternateTree

lossStatistics

protected DoubleVector lossStatistics

weightSeen

protected int weightSeen

previousWeight

protected int previousWeight

Constructor Detail

FIMTDD.FIMTDDSplitNode

public FIMTDD.FIMTDDSplitNode(InstanceConditionalTest splitTest,
                      double[] classObservations)

Create a new FIMTDDSplitNode

Method Detail

setParent

public void setParent(HoeffdingTree.Node parent)

Set the parent node

Specified by:

setParent in interface FIMTDD.AdaptationCompatibleNode

getParent

public HoeffdingTree.Node getParent()

Return the parent node

Specified by:

getParent in interface FIMTDD.AdaptationCompatibleNode

PageHinckleyTest

public boolean PageHinckleyTest(double error,
                       double threshold)

Check to see if the tree needs updating

learnFromInstance

public void learnFromInstance(weka.core.Instance inst,
                     FIMTDD ht,
                     boolean growingAltTree)

This method has been added to the SplitNode subclass to allow alternate trees to be grown and compared to the current tree, to allow adaptation in the face of concept drift