org.apache.spark.mllib.tree.model

Class Bin

Object

org.apache.spark.mllib.tree.model.Bin

All Implemented Interfaces:

java.io.Serializable, scala.Equals, scala.Product

public class Bin
extends Object
implements scala.Product, scala.Serializable

Used for "binning" the feature values for faster best split calculation.

For a continuous feature, the bin is determined by a low and a high split, where an example with featureValue falls into the bin s.t. lowSplit.threshold < featureValue <= highSplit.threshold.

For ordered categorical features, there is a 1-1-1 correspondence between bins, splits, and feature values. The bin is determined by category/feature value. However, the bins are not necessarily ordered by feature value; they are ordered using impurity.

For unordered categorical features, there is a 1-1 correspondence between bins, splits, where bins and splits correspond to subsets of feature values (in highSplit.categories). An unordered feature with k categories uses (1 << k - 1) - 1 bins, corresponding to all partitionings of categories into 2 disjoint, non-empty sets.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
Bin(Split lowSplit, Split highSplit, scala.Enumeration.Value featureType, double category)

Method Summary

Methods

Modifier and Type	Method and Description
double	category()
scala.Enumeration.Value	featureType()
Split	highSplit()
Split	lowSplit()

Methods inherited from class Object

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

Methods inherited from interface scala.Product

productArity, productElement, productIterator, productPrefix

Methods inherited from interface scala.Equals

canEqual, equals

Constructor Detail

Bin

public Bin(Split lowSplit,
   Split highSplit,
   scala.Enumeration.Value featureType,
   double category)

Method Detail

lowSplit

public Split lowSplit()

highSplit

public Split highSplit()

featureType

public scala.Enumeration.Value featureType()

category

public double category()