org.apache.spark.mllib.tree.impurity

Class ImpurityCalculator

Object

org.apache.spark.mllib.tree.impurity.ImpurityCalculator

Direct Known Subclasses:

EntropyCalculator, GiniCalculator, VarianceCalculator

public abstract class ImpurityCalculator
extends Object

Stores statistics for one (node, feature, bin) for calculating impurity. Unlike ImpurityAggregator, this class stores its own data and is for a specific (node, feature, bin).

Constructor Summary

Constructors

Constructor and Description
ImpurityCalculator(double[] stats)

Method Summary

Methods

Modifier and Type	Method and Description
ImpurityCalculator	add(ImpurityCalculator other) Add the stats from another calculator into this one, modifying and returning this calculator.
abstract double	calculate() Calculate the impurity from the stored sufficient statistics.
abstract ImpurityCalculator	copy() Make a deep copy of this ImpurityCalculator.
abstract long	count() Number of data points accounted for in the sufficient statistics.
abstract double	predict() Prediction which should be made based on the sufficient statistics.
double	prob(double label) Probability of the label given by predict, or -1 if no probability is available.
double[]	stats()
ImpurityCalculator	subtract(ImpurityCalculator other) Subtract the stats from another calculator from this one, modifying and returning this calculator.

Methods inherited from class Object

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

Constructor Detail

ImpurityCalculator

public ImpurityCalculator(double[] stats)

Method Detail

stats

public double[] stats()

copy

public abstract ImpurityCalculator copy()

Make a deep copy of this ImpurityCalculator.

calculate

public abstract double calculate()

Calculate the impurity from the stored sufficient statistics.

add

public ImpurityCalculator add(ImpurityCalculator other)

Add the stats from another calculator into this one, modifying and returning this calculator.

subtract

public ImpurityCalculator subtract(ImpurityCalculator other)

Subtract the stats from another calculator from this one, modifying and returning this calculator.

count

public abstract long count()

Number of data points accounted for in the sufficient statistics.

predict

public abstract double predict()

Prediction which should be made based on the sufficient statistics.

prob

public double prob(double label)

Probability of the label given by predict, or -1 if no probability is available.