weka.classifiers.mi.miti

Class Gini

java.lang.Object

weka.classifiers.mi.miti.Gini

All Implemented Interfaces:

IBestSplitMeasure

public class Gini
extends java.lang.Object
implements IBestSplitMeasure

Implements the Gini-based split selection measure.

Version:

$Revision: 8109 $

Author:

Luke Bjerring

Constructor Summary

Constructors

Constructor and Description
Gini()

Method Summary

Modifier and Type	Method and Description
static double	getGiniImpurity(double[] totalCounts, double[] positiveCounts, int kBEPPConstant, boolean unbiasedEstimate) Returns Gini impurity score for the N groups after a nominal split (the closer to zero, the better the split).
static double	getGiniImpurity(SufficientStatistics ss, int kBEPPConstant, boolean unbiasedEstimate) Returns Gini impurity score of the two groups after the split - the closer to zero, the better the split.
double	getScore(double[] totalCounts, double[] positiveCounts, int kBEPPConstant, boolean unbiasedEstimate) Returns purity score for the N groups after a nominal split - larger is better
double	getScore(SufficientStatistics ss, int kBEPPConstant, boolean unbiasedEstimate) Returns purity score of the two groups after the split - larger is better

Methods inherited from class java.lang.Object

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

Constructor Detail

Gini

public Gini()

Method Detail

getGiniImpurity

public static double getGiniImpurity(SufficientStatistics ss,
                                     int kBEPPConstant,
                                     boolean unbiasedEstimate)

Returns Gini impurity score of the two groups after the split - the closer to zero, the better the split.

getGiniImpurity

public static double getGiniImpurity(double[] totalCounts,
                                     double[] positiveCounts,
                                     int kBEPPConstant,
                                     boolean unbiasedEstimate)

Returns Gini impurity score for the N groups after a nominal split (the closer to zero, the better the split).

getScore

public double getScore(SufficientStatistics ss,
                       int kBEPPConstant,
                       boolean unbiasedEstimate)

Returns purity score of the two groups after the split - larger is better

Specified by:

getScore in interface IBestSplitMeasure

getScore

public double getScore(double[] totalCounts,
                       double[] positiveCounts,
                       int kBEPPConstant,
                       boolean unbiasedEstimate)

Returns purity score for the N groups after a nominal split - larger is better

Specified by:

getScore in interface IBestSplitMeasure