com.bayesia.api.discretization.settings

Class KMeansSettings

java.lang.Object

com.bayesia.api.discretization.settings.KMeansSettings

All Implemented Interfaces:

DiscretizationSettings

public class KMeansSettings
extends java.lang.Object
implements DiscretizationSettings

This class is used to perform discretization using the K-Means discretization algorithm. It must be used with APIUtils.discretize(DiscretizationSettings, double[], double[])

Version:

4.0

Author:

Bayesia S.A.S.

Constructor Summary

Constructors

Constructor and Description
KMeansSettings(int intervalCount, boolean logUsed, boolean zeroIsolated) Create settings for the K-Means discretization algorithm providing the number of intervals we want to obtain.

Method Summary

Modifier and Type	Method and Description
APIUtils.DISCRETIZATION	getDiscretization() Return a DISCRETIZATION indicating the discretization algorithm to use
int	getIntervalCount()
boolean	isLogUsed()
boolean	isZeroIsolated()

Methods inherited from class java.lang.Object

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

Constructor Detail

KMeansSettings

public KMeansSettings(int intervalCount,
                      boolean logUsed,
                      boolean zeroIsolated)

Create settings for the K-Means discretization algorithm providing the number of intervals we want to obtain. This number must be superior or equal to 2. In some cases of data distribution, it is possible to use the log10 of the data before performing discretization by specifying logUsed. When there is a lot of zeros in the data, it is possible to consider them as a unique zero, avoiding discretization issues and focusing discretization process over the non-zero values.

Parameters:

intervalCount - the number of intervals

logUsed - indicates if the data have to be converted into log before discretization

zeroIsolated - indicates if we want to consider all zeros as a unique value

Method Detail

getDiscretization

public APIUtils.DISCRETIZATION getDiscretization()

Description copied from interface: DiscretizationSettings

Return a DISCRETIZATION indicating the discretization algorithm to use

Specified by:

getDiscretization in interface DiscretizationSettings

Returns:

the discretization to use

getIntervalCount

public int getIntervalCount()

Returns:

the number of intervals

isLogUsed

public boolean isLogUsed()

Returns:

true if the data have to be converted as log before discretization, false otherwise

isZeroIsolated

public boolean isZeroIsolated()

Returns:

true if all zeros are treated as a unique value