weka.classifiers.trees.ht

Class SplitMetric

java.lang.Object

weka.classifiers.trees.ht.SplitMetric

All Implemented Interfaces:

java.io.Serializable

Direct Known Subclasses:

GiniSplitMetric, InfoGainSplitMetric

public abstract class SplitMetric
extends java.lang.Object
implements java.io.Serializable

Base class for split metrics

Version:

$Revision: 9720 $

Author:

Richard Kirkby (rkirkby@cs.waikato.ac.nz), Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
SplitMetric()

Method Summary

Modifier and Type	Method and Description
abstract double	evaluateSplit(java.util.Map<java.lang.String,WeightMass> preDist, java.util.List<java.util.Map<java.lang.String,WeightMass>> postDist) Evaluate the merit of a split
abstract double	getMetricRange(java.util.Map<java.lang.String,WeightMass> preDist) Get the range of the splitting metric
static double	sum(java.util.Map<java.lang.String,WeightMass> dist) Utility method to return the sum of instance weight in a distribution

Methods inherited from class java.lang.Object

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

Constructor Detail

SplitMetric

public SplitMetric()

Method Detail

sum

public static double sum(java.util.Map<java.lang.String,WeightMass> dist)

Utility method to return the sum of instance weight in a distribution

Parameters:

dist - the distribution

Returns:

the sum of the weights contained in a distribution

evaluateSplit

public abstract double evaluateSplit(java.util.Map<java.lang.String,WeightMass> preDist,
                                     java.util.List<java.util.Map<java.lang.String,WeightMass>> postDist)

Evaluate the merit of a split

Parameters:

preDist - the class distribution before the split

postDist - the class distributions after the split

Returns:

the merit of the split

getMetricRange

public abstract double getMetricRange(java.util.Map<java.lang.String,WeightMass> preDist)

Get the range of the splitting metric

Parameters:

preDist - the pre-split class distribution

Returns:

the range of the splitting metric