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

Class ImpurityHistogram

java.lang.Object

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

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

GiniHistogram, MSEHistogram

public abstract class ImpurityHistogram
extends Object
implements Serializable

Helper class for ImpurityHistograms.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected Set<Integer>	bucketIds Bucket ids.
protected int	featureId Feature id.

Constructor Summary

Constructors

Constructor and Description
ImpurityHistogram(int featureId) Creates an instance of ImpurityHistogram.

Method Summary

Modifier and Type	Method and Description
protected Optional<NodeSplit>	checkAndReturnSplitValue(int bestBucketId, double bestSplitVal, double bestGain, double impurity) Checks split value validity and return Optional-wrap of it.

Methods inherited from class java.lang.Object

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

Field Detail

bucketIds

protected final Set<Integer> bucketIds

Bucket ids.

featureId

protected final int featureId

Feature id.

Constructor Detail

ImpurityHistogram

public ImpurityHistogram(int featureId)

Creates an instance of ImpurityHistogram.

Parameters:

featureId - Feature id.

Method Detail

checkAndReturnSplitValue

protected Optional<NodeSplit> checkAndReturnSplitValue(int bestBucketId,
                                                       double bestSplitVal,
                                                       double bestGain,
                                                       double impurity)

Checks split value validity and return Optional-wrap of it. In other case returns Optional.empty

Parameters:

bestBucketId - Best bucket id.

bestSplitVal - Best split value.

bestGain - Best gain.

impurity - Node's impurity.

Returns:

Best split value.