Package elki.outlier.spatial

Class CTLuGLSBackwardSearchAlgorithm<V extends NumberVector>

java.lang.Object

elki.outlier.spatial.CTLuGLSBackwardSearchAlgorithm<V>

Type Parameters:

V - Vector type to use for distances

All Implemented Interfaces:

Algorithm, OutlierAlgorithm

@Title("GLS-Backward Search")
@Reference(authors="F. Chen, C.-T. Lu, A. P. Boedihardjo",
           title="GLS-SOD: A Generalized Local Statistical Approach for Spatial Outlier Detection",
           booktitle="Proc. 16th ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining",
           url="https://doi.org/10.1145/1835804.1835939",
           bibkey="DBLP:conf/kdd/ChenLB10")
public class CTLuGLSBackwardSearchAlgorithm<V extends NumberVector>
extends java.lang.Object
implements OutlierAlgorithm

GLS-Backward Search is a statistical approach to detecting spatial outliers.

Implementation note: this is just the most basic version of this algorithm. The spatial relation must be two dimensional, the set of spatial basis functions is hard-coded (but trivial to enhance) to \(\{1,x,y,x^2,y^2,xy\}\), and we assume the neighborhood is large enough for the simpler formulas to work that make the optimization problem convex.

Reference:

F. Chen, C.-T. Lu, A. P. Boedihardjo
GLS-SOD: A Generalized Local Statistical Approach for Spatial Outlier Detection
Proc. 16th ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining

Since:

0.4.0

Author:

Ahmed Hettab

Nested Class Summary

Nested classes/interfaces inherited from interface elki.Algorithm

Algorithm.Utils

Field Summary

Fields

Modifier and Type	Field	Description
protected double	alpha	Parameter Alpha - significance niveau
protected Distance<? super V>	distance	Distance function used.
protected int	k	Parameter k - neighborhood size

Constructor Summary

Constructors

Constructor	Description
CTLuGLSBackwardSearchAlgorithm(Distance<? super V> distance, int k, double alpha)	Constructor.

Method Summary

Modifier and Type	Method	Description
TypeInformation[]	getInputTypeRestriction()	Get the input type restriction used for negotiating the data query.
OutlierResult	run(Relation<V> relationx, Relation<? extends NumberVector> relationy)	Run the algorithm
private Pair<DBIDVar,java.lang.Double>	singleIteration(Relation<V> relationx, Relation<? extends NumberVector> relationy)	Run a single iteration of the GLS-SOD modeling step

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

distance

protected Distance<? super V extends NumberVector> distance

Distance function used.

alpha

protected double alpha

Parameter Alpha - significance niveau

k

protected int k

Parameter k - neighborhood size

Constructor Detail

CTLuGLSBackwardSearchAlgorithm

public CTLuGLSBackwardSearchAlgorithm(Distance<? super V> distance,
                                      int k,
                                      double alpha)

Constructor.

Parameters:

distance - Distance function

k - number of nearest neighbors to use

alpha - Significance niveau

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<V> relationx,
                         Relation<? extends NumberVector> relationy)

Run the algorithm

Parameters:

relationx - Spatial relation

relationy - Attribute relation

Returns:

Algorithm result

singleIteration

private Pair<DBIDVar,java.lang.Double> singleIteration(Relation<V> relationx,
                                                             Relation<? extends NumberVector> relationy)

Run a single iteration of the GLS-SOD modeling step

Parameters:

relationx - Geo relation

relationy - Attribute relation

Returns:

Top outlier and associated score