org.apache.ignite.ml.tree.randomforest.data.impurity

Class GiniHistogram

java.lang.Object

org.apache.ignite.ml.tree.randomforest.data.impurity.ImpurityHistogram

org.apache.ignite.ml.tree.randomforest.data.impurity.GiniHistogram

All Implemented Interfaces:

Serializable, Histogram<BootstrappedVector,GiniHistogram>, ImpurityComputer<BootstrappedVector,GiniHistogram>

public class GiniHistogram
extends ImpurityHistogram
implements ImpurityComputer<BootstrappedVector,GiniHistogram>

Class contains implementation of splitting point finding algorithm based on Gini metric (see https://en.wikipedia.org/wiki/Gini_coefficient) and represents a set of histograms in according to this metric.

See Also:

Serialized Form

Field Summary

Fields inherited from class org.apache.ignite.ml.tree.randomforest.data.impurity.ImpurityHistogram

featureId

Constructor Summary

Constructors

Constructor and Description
GiniHistogram(int sampleId, Map<Double,Integer> lblMapping, BucketMeta bucketMeta) Creates an instance of GiniHistogram.

Method Summary

Modifier and Type	Method and Description
void	addElement(BootstrappedVector vector) Add object to histogram.
Set<Integer>	buckets()
Optional<NodeSplit>	findBestSplit() Returns best split point computed on histogram if it exists.
Optional<Double>	getValue(Integer bucketId)
boolean	isEqualTo(GiniHistogram other) Compares histogram with other and returns true if they are equals
GiniHistogram	plus(GiniHistogram other)

Methods inherited from class org.apache.ignite.ml.tree.randomforest.data.impurity.ImpurityHistogram

checkAndReturnSplitValue

Methods inherited from class java.lang.Object

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

Constructor Detail

GiniHistogram

public GiniHistogram(int sampleId,
                     Map<Double,Integer> lblMapping,
                     BucketMeta bucketMeta)

Creates an instance of GiniHistogram.

Parameters:

sampleId - Sample id.

lblMapping - Label mapping.

bucketMeta - Bucket meta.

Method Detail

addElement

public void addElement(BootstrappedVector vector)

Add object to histogram.

Specified by:

addElement in interface Histogram<BootstrappedVector,GiniHistogram>

Parameters:

vector - Value.

getValue

public Optional<Double> getValue(Integer bucketId)

Specified by:

getValue in interface Histogram<BootstrappedVector,GiniHistogram>

Parameters:

bucketId - Bucket id.

Returns:

Value in according to bucket id.

plus

public GiniHistogram plus(GiniHistogram other)

Specified by:

plus in interface Histogram<BootstrappedVector,GiniHistogram>

Parameters:

other - Other histogram.

Returns:

Sum of this and other histogram.

findBestSplit

public Optional<NodeSplit> findBestSplit()

Returns best split point computed on histogram if it exists. Split point may be absent when there is no data in histograms or split point lay in last bucket in histogram.

Specified by:

findBestSplit in interface ImpurityComputer<BootstrappedVector,GiniHistogram>

Returns:

Splitting point for decision tree.

buckets

public Set<Integer> buckets()

Specified by:

buckets in interface Histogram<BootstrappedVector,GiniHistogram>

Returns:

Bucket ids.

isEqualTo

public boolean isEqualTo(GiniHistogram other)

Compares histogram with other and returns true if they are equals

Specified by:

isEqualTo in interface Histogram<BootstrappedVector,GiniHistogram>

Parameters:

other - Other histogram.

Returns:

True if histograms are equal.