Package elki.outlier.lof

Class VarianceOfVolume<O extends SpatialComparable>

java.lang.Object

elki.outlier.lof.VarianceOfVolume<O>

Type Parameters:

O - the type of data objects handled by this algorithm

All Implemented Interfaces:

Algorithm, OutlierAlgorithm

@Reference(authors="T. Hu, S. Y. Sung",
           title="Detecting pattern-based outliers",
           booktitle="Pattern Recognition Letters 24(16)",
           url="https://doi.org/10.1016/S0167-8655(03)00165-X",
           bibkey="DBLP:journals/prl/HuS03")
public class VarianceOfVolume<O extends SpatialComparable>
extends java.lang.Object
implements OutlierAlgorithm

Variance of Volume for outlier detection.

The volume is estimated by the distance to the k-nearest neighbor, then the variance of volume is computed.

Unfortunately, this approach needs an enormous numerical precision, and may not work for high-dimensional, non-normalized data. We therefore divide each volume by the average across the data set. This means values are even less comparable across data sets, but this avoids some of the numerical problems of this method.

Reference:

T. Hu, S. Y. Sung
Detecting pattern-based outliers
Pattern Recognition Letters 24(16)

Since:

0.7.0

Author:

Erich Schubert

Nested Class Summary

Nested classes/interfaces inherited from interface elki.Algorithm

Algorithm.Utils

Field Summary

Fields

Modifier and Type	Field	Description
protected Distance<? super O>	distance	Distance function used.
protected int	kplus	The number of neighbors to query (plus the query point!)
private static Logging	LOG	The logger for this class.

Constructor Summary

Constructors

Constructor	Description
VarianceOfVolume(int k, Distance<? super O> distance)	Constructor.

Method Summary

Modifier and Type	Method	Description
private void	computeVolumes(KNNSearcher<DBIDRef> knnq, int dim, DBIDs ids, WritableDoubleDataStore vols)	Compute volumes
private void	computeVOVs(KNNSearcher<DBIDRef> knnq, DBIDs ids, DoubleDataStore vols, WritableDoubleDataStore vovs, DoubleMinMax vovminmax)	Compute variance of volumes.
TypeInformation[]	getInputTypeRestriction()	Get the input type restriction used for negotiating the data query.
OutlierResult	run(Relation<O> relation)	Runs the VOV algorithm on the given database.

Methods inherited from class java.lang.Object

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

Methods inherited from interface elki.outlier.OutlierAlgorithm

autorun

Field Detail

LOG

private static final Logging LOG

The logger for this class.

distance

protected Distance<? super O extends SpatialComparable> distance

Distance function used.

kplus

protected int kplus

The number of neighbors to query (plus the query point!)

Constructor Detail

VarianceOfVolume

public VarianceOfVolume(int k,
                        Distance<? super O> distance)

Constructor.

Parameters:

k - number of neighbors to use for comparison

distance - the neighborhood distance function

Method Detail

getInputTypeRestriction

public TypeInformation[] getInputTypeRestriction()

Description copied from interface: Algorithm

Get the input type restriction used for negotiating the data query.

Specified by:

getInputTypeRestriction in interface Algorithm

Returns:

Type restriction

run

public OutlierResult run(Relation<O> relation)

Runs the VOV algorithm on the given database.

Parameters:

relation - Data to process

Returns:

VOV outlier result

computeVolumes

private void computeVolumes(KNNSearcher<DBIDRef> knnq,
                            int dim,
                            DBIDs ids,
                            WritableDoubleDataStore vols)

Compute volumes

Parameters:

knnq - KNN query

dim - Data dimensionality

ids - IDs to process

vols - Volume storage

computeVOVs

private void computeVOVs(KNNSearcher<DBIDRef> knnq,
                         DBIDs ids,
                         DoubleDataStore vols,
                         WritableDoubleDataStore vovs,
                         DoubleMinMax vovminmax)

Compute variance of volumes.

Parameters:

knnq - KNN query

ids - IDs to process

vols - Volumes

vovs - Variance of Volume storage

vovminmax - Score minimum/maximum tracker