hex.tree.isoforextended.isolationtree

Class IsolationTree

java.lang.Object

hex.tree.isoforextended.isolationtree.IsolationTree

public class IsolationTree
extends java.lang.Object

IsolationTree class implements Algorithm 2 (iTree) Naming convention comes from the Extended Isolation Forest paper.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	IsolationTree.FilteredData
static class	IsolationTree.Node IsolationTree Node.

Constructor Summary

Constructors

Constructor and Description
IsolationTree(int _heightLimit, int _extensionLevel)

Method Summary

Modifier and Type	Method and Description
CompressedIsolationTree	buildTree(double[][] data, long seed, int treeNum) Implementation of Algorithm 2 (iTree) from paper.
static IsolationTree.FilteredData	extendedIsolationForestSplit(double[][] data, double[] p, double[] n) Compute Extended Isolation Forest split point and filter input data with this split point in the same time.
static double[]	gaussianVector(int n, int zeroNum, long seed) Make a new array initialized to random Gaussian N(0,1) values with the given seed.
int	getDepth()
int	getIsolatedPoints()
int	getLeaves()
long	getNotIsolatedPoints()
int	getZeroSplits()
void	logNodesHeight(org.apache.log4j.Level level) Helper method.
void	logNodesNumRows(org.apache.log4j.Level level) Helper method.

Methods inherited from class java.lang.Object

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

Constructor Detail

IsolationTree

public IsolationTree(int _heightLimit,
                     int _extensionLevel)

Method Detail

buildTree

public CompressedIsolationTree buildTree(double[][] data,
                                         long seed,
                                         int treeNum)

Implementation of Algorithm 2 (iTree) from paper.

logNodesNumRows

public void logNodesNumRows(org.apache.log4j.Level level)

Helper method. Print nodes' size of the tree.

logNodesHeight

public void logNodesHeight(org.apache.log4j.Level level)

Helper method. Print height (length of path from root) of each node in trees. Root is 0.

extendedIsolationForestSplit

public static IsolationTree.FilteredData extendedIsolationForestSplit(double[][] data,
                                                                      double[] p,
                                                                      double[] n)

Compute Extended Isolation Forest split point and filter input data with this split point in the same time.

See Algorithm 2 (iTree) in the paper.

Returns:

Object containing data for Left and Right branch of the tree.

gaussianVector

public static double[] gaussianVector(int n,
                                      int zeroNum,
                                      long seed)

Make a new array initialized to random Gaussian N(0,1) values with the given seed. Make randomly selected zeroNum items zeros (based on extensionLevel value).

Parameters:

n - length of generated vector

zeroNum - set randomly selected zeroNum items of vector to zero

Returns:

array with gaussian values. Randomly selected zeroNum item values are zeros.

getIsolatedPoints

public int getIsolatedPoints()

getNotIsolatedPoints

public long getNotIsolatedPoints()

getZeroSplits

public int getZeroSplits()

getLeaves

public int getLeaves()

getDepth

public int getDepth()