Main Page | Class Hierarchy | Class List | Class Members

kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure Class Reference

Inheritance diagram for kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure:

kddml.Core.DataMining.Clustering.ComparisonMeasure kddml.Core.DataMining.Clustering.BINARY_SIMILARITY kddml.Core.DataMining.Clustering.CHEBYCHEV kddml.Core.DataMining.Clustering.CITY_BLOCK kddml.Core.DataMining.Clustering.EUCLIDEAN kddml.Core.DataMining.Clustering.JACCARD kddml.Core.DataMining.Clustering.MINKOWSKI kddml.Core.DataMining.Clustering.SIMPLE_MATCHING kddml.Core.DataMining.Clustering.SQUARED_EUCLIDEAN kddml.Core.DataMining.Clustering.TANIMOTO List of all members.

Public Member Functions

 CentroidBasedComparisonMeasure (AttributeComparisonMeasure[] attribute_comparison, double[] weights)
 CentroidBasedComparisonMeasure (ClusterDescription cluster_description)
abstract boolean isSimilarityMeasure ()
double compare (ClusterManager cluster, Instance instance) throws ClusteringModelException
boolean isCentroidBased ()
AttributeComparisonMeasure[] getAttributeComparisonMeasure ()

Protected Member Functions

abstract double evaluate (double[] x)
Element toXML ()

Detailed Description

A comparison measure centroid-based. A centroid-based comparison measure can be used either for a centroid-based clustering and a distribution-based clustering.

Title: KDDML

Description: Knowledge Discovery in Database Environment

Copyright: Copyright (c) 2003-2005

Company: Universita' di Pisa - Dipartimento di Informatica

Author:
Andrea Romei (romei@di.unipi.it)
Version:
2.0.16


Constructor & Destructor Documentation

kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.CentroidBasedComparisonMeasure AttributeComparisonMeasure[]  attribute_comparison,
double[]  weights
 

Constructor given the list of comparison measures. Used for centroid-based clustering only.

Parameters:
attribute_comparison AttributeComparisonMeasure[] the comparison measure for each attribute.
weights double[] the field weight for each attribute. Can be null.

kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.CentroidBasedComparisonMeasure ClusterDescription  cluster_description  ) 
 

Constructor given the centroid description. Used both for centroid-based clustering and distribution-based clustering.

Parameters:
cluster_description ClusterDescription


Member Function Documentation

abstract boolean kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.isSimilarityMeasure  )  [pure virtual]
 

Returns true if the comparison measure is a similarity function, in which the value returned by the compare method is optimal for greater values. Returns false if the comparison measure is a distance measure, in which the value returned by the compare method is optimal if it is 0.

Returns:
boolean

Implemented in kddml.Core.DataMining.Clustering.BINARY_SIMILARITY, kddml.Core.DataMining.Clustering.CHEBYCHEV, kddml.Core.DataMining.Clustering.CITY_BLOCK, kddml.Core.DataMining.Clustering.EUCLIDEAN, kddml.Core.DataMining.Clustering.JACCARD, kddml.Core.DataMining.Clustering.MINKOWSKI, kddml.Core.DataMining.Clustering.SIMPLE_MATCHING, kddml.Core.DataMining.Clustering.SQUARED_EUCLIDEAN, and kddml.Core.DataMining.Clustering.TANIMOTO.

double kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.compare ClusterManager  cluster,
Instance  instance
throws ClusteringModelException
 

Compares an input istance with the seed of the cluster. The seed depends on the type of clustering. If the clustering is center-based, the seed is the centroid as instance. If the clustering is distribution-based, the seed is calculated on the statistics associated to the cluster. In particular, for numeric attributes, the mean of cluster instances is used as seed. For discrete attribute, the most probable category is reported as seed.

Parameters:
cluster ClusterManager
instance Instance
Exceptions:
ClusteringModelException 
Returns:
double the optimal values depends on the isSimilarityMeasure method.

Reimplemented in kddml.Core.DataMining.Clustering.SIMPLE_MATCHING.

abstract double kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.evaluate double[]  x  )  [protected, pure virtual]
 

Evaluates the comparison measure given the comparison values for each single attribute (i.e., the values returned by each single inner function). This method is implemented in sub-classes.

Parameters:
x double[]
Returns:
double

Implemented in kddml.Core.DataMining.Clustering.BINARY_SIMILARITY, kddml.Core.DataMining.Clustering.CHEBYCHEV, kddml.Core.DataMining.Clustering.CITY_BLOCK, kddml.Core.DataMining.Clustering.EUCLIDEAN, kddml.Core.DataMining.Clustering.JACCARD, kddml.Core.DataMining.Clustering.MINKOWSKI, kddml.Core.DataMining.Clustering.SIMPLE_MATCHING, kddml.Core.DataMining.Clustering.SQUARED_EUCLIDEAN, and kddml.Core.DataMining.Clustering.TANIMOTO.

boolean kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.isCentroidBased  )  [virtual]
 

Returns true if the function is related to a centroid-based clustering. Returns true if the function is related to a distribution-based clustering.

Returns:
boolean true

Implements kddml.Core.DataMining.Clustering.ComparisonMeasure.

AttributeComparisonMeasure [] kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.getAttributeComparisonMeasure  ) 
 

Returns the comparison measure for each attribute.

Returns:
AttributeComparisonMeasure[]

Element kddml.Core.DataMining.Clustering.CentroidBasedComparisonMeasure.toXML  )  [protected, virtual]
 

Returns a representation of this function as PMML element.

Returns:
Element

Implements kddml.Core.DataMining.Clustering.ComparisonMeasure.


Generated on Thu Feb 23 13:04:40 2006 for kddml by  doxygen 1.4.3