weka.clusterers
Class Cobweb.CNode

java.lang.Object
  weka.clusterers.Cobweb.CNode

All Implemented Interfaces:

java.io.Serializable

Enclosing class:

Cobweb

private class Cobweb.CNode

extends java.lang.Object

implements java.io.Serializable

Inner class handling node operations for Cobweb.

See Also:

Serializable

Field Summary

private AttributeStats[]	m_attStats Within cluster attribute statistics
private FastVector	m_children Children of this node
protected Instances	m_clusterInstances Instances at this node
private int	m_clusterNum Cluster number of this node
private int	m_numAttributes Number of attributes
private double	m_totalInstances Total instances at this node

Constructor Summary

Cobweb.CNode(int numAttributes)
Creates an empty CNode instance.

Cobweb.CNode(int numAttributes, Instance leafInstance)
Creates a new leaf CNode instance.

Method Summary

protected void	addChildNode(Cobweb.CNode child) Adds the supplied node as a child of this node.
protected void	addInstance(Instance newInstance) Adds an instance to this cluster.
private void	assignClusterNums(int[] cl_num) Recursively assigns numbers to the nodes in the tree.
protected double	categoryUtility() Computes the utility of all children with respect to this node
protected double	categoryUtilityChild(Cobweb.CNode child) Computes the utility of a single child with respect to this node
private double	cuScoreForBestTwoMerged(Cobweb.CNode merged, Cobweb.CNode a, Cobweb.CNode b, Instance newInstance)
private double[]	cuScoresForChildren(Instance newInstance) Temporarily adds a new instance to each of this nodes children in turn and computes the category utility.
protected java.lang.String	dumpData() Returns the instances at this node as a string.
protected void	dumpTree(int depth, java.lang.StringBuffer text) Recursively build a string representation of the Cobweb tree
private Cobweb.CNode	findHost(Instance newInstance, boolean structureFrozen) Finds a host for the new instance in this nodes children.
protected double	getProbability(int attIndex, int valueIndex) Returns the probability of a value of a nominal attribute in this node
protected double	getStandardDev(int attIndex) Returns the standard deviation of a numeric attribute
protected void	graphTree(java.lang.StringBuffer text) Recursively generate the graph string for the Cobweb tree.
protected void	updateStats(Instance updateInstance, boolean delete) Update attribute stats using the supplied instance.

Methods inherited from class java.lang.Object

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

Field Detail

m_attStats

private AttributeStats[] m_attStats

Within cluster attribute statistics

m_numAttributes

private int m_numAttributes

Number of attributes

m_clusterInstances

protected Instances m_clusterInstances

Instances at this node

m_children

private FastVector m_children

Children of this node

m_totalInstances

private double m_totalInstances

Total instances at this node

m_clusterNum

private int m_clusterNum

Cluster number of this node

Constructor Detail

Cobweb.CNode

public Cobweb.CNode(int numAttributes)

Creates an empty CNode instance.

Parameters:

numAttributes - the number of attributes in the data

Cobweb.CNode

public Cobweb.CNode(int numAttributes,
                    Instance leafInstance)

Creates a new leaf CNode instance.

Parameters:

numAttributes - the number of attributes in the data

leafInstance - the instance to store at this leaf

Method Detail

addInstance

protected void addInstance(Instance newInstance)
                    throws java.lang.Exception

Adds an instance to this cluster.

Parameters:

newInstance - the instance to add

Throws:

java.lang.Exception - if an error occurs

cuScoresForChildren

private double[] cuScoresForChildren(Instance newInstance)
                              throws java.lang.Exception

Temporarily adds a new instance to each of this nodes children in turn and computes the category utility.

Parameters:

newInstance - the new instance to evaluate

Returns:

an array of category utility values---the result of considering each child in turn as a host for the new instance

Throws:

java.lang.Exception - if an error occurs

cuScoreForBestTwoMerged

private double cuScoreForBestTwoMerged(Cobweb.CNode merged,
                                       Cobweb.CNode a,
                                       Cobweb.CNode b,
                                       Instance newInstance)
                                throws java.lang.Exception

Throws:

java.lang.Exception

findHost

private Cobweb.CNode findHost(Instance newInstance,
                              boolean structureFrozen)
                       throws java.lang.Exception

Finds a host for the new instance in this nodes children. Also considers merging the two best hosts and splitting the best host.

Parameters:

newInstance - the instance to find a host for

structureFrozen - true if the instance is not to be added to the tree and instead the best potential host is to be returned

Returns:

the best host

Throws:

java.lang.Exception - if an error occurs

addChildNode

protected void addChildNode(Cobweb.CNode child)

Adds the supplied node as a child of this node. All of the child's instances are added to this nodes instances

Parameters:

child - the child to add

categoryUtility

protected double categoryUtility()
                          throws java.lang.Exception

Computes the utility of all children with respect to this node

Returns:

the category utility of the children with respect to this node.

Throws:

java.lang.Exception

categoryUtilityChild

protected double categoryUtilityChild(Cobweb.CNode child)
                               throws java.lang.Exception

Computes the utility of a single child with respect to this node

Parameters:

child - the child for which to compute the utility

Returns:

the utility of the child with respect to this node

Throws:

java.lang.Exception - if something goes wrong

getProbability

protected double getProbability(int attIndex,
                                int valueIndex)
                         throws java.lang.Exception

Returns the probability of a value of a nominal attribute in this node

Parameters:

attIndex - the index of the attribute

valueIndex - the index of the value of the attribute

Returns:

the probability

Throws:

java.lang.Exception - if the requested attribute is not nominal

getStandardDev

protected double getStandardDev(int attIndex)
                         throws java.lang.Exception

Returns the standard deviation of a numeric attribute

Parameters:

attIndex - the index of the attribute

Returns:

the standard deviation

Throws:

java.lang.Exception - if an error occurs

updateStats

protected void updateStats(Instance updateInstance,
                           boolean delete)

Update attribute stats using the supplied instance.

Parameters:

updateInstance - the instance for updating

delete - true if the values of the supplied instance are to be removed from the statistics

assignClusterNums

private void assignClusterNums(int[] cl_num)
                        throws java.lang.Exception

Recursively assigns numbers to the nodes in the tree.

Parameters:

cl_num - an int[] value

Throws:

java.lang.Exception - if an error occurs

dumpTree

protected void dumpTree(int depth,
                        java.lang.StringBuffer text)

Recursively build a string representation of the Cobweb tree

Parameters:

depth - depth of this node in the tree

text - holds the string representation

dumpData

protected java.lang.String dumpData()
                             throws java.lang.Exception

Returns the instances at this node as a string. Appends the cluster number of the child that each instance belongs to.

Returns:

a String value

Throws:

java.lang.Exception - if an error occurs

graphTree

protected void graphTree(java.lang.StringBuffer text)
                  throws java.lang.Exception

Recursively generate the graph string for the Cobweb tree.

Parameters:

text - holds the graph string

Throws:

java.lang.Exception