org.jpmml.evaluator

Interface HasAffinity

All Superinterfaces:

CategoricalResultFeature, ResultFeature

All Known Subinterfaces:

HasAffinityRanking

All Known Implementing Classes:

ClusterClassificationMap, InstanceClassificationMap

public interface HasAffinity
extends CategoricalResultFeature

Affinity represents a degree of attraction between the sample and a particular category.

PMML deals with two kinds of affinities:

• Distance between two points in an n-dimensional feature space. Smaller distance values indicate more optimal fit.
• Similarity between two feature vectors. Greater similarity values indicate more optimal fit.

Clustering is an unsupervised learning task. The set of all categories is defined by the identifiers of reference entities. For clustering models this includes all clusters. For k-nearest neighbor models this includes k most optimal training instances.

See Also:

ResultFeatureType.AFFINITY

Method Summary

Methods

Modifier and Type	Method and Description
Double	getAffinity(String value) Gets the affinity towards the specified category.

Methods inherited from interface org.jpmml.evaluator.CategoricalResultFeature

getCategoryValues

Method Detail

getAffinity

Double getAffinity(String value)

Gets the affinity towards the specified category.

Returns:

An affinity in the range from 0.0 to positive infinity. The affinity of an unknown category is the least optimal value in the range of valid values (ie. Double.POSITIVE_INFINITY for distance measures and 0.0 for similarity measures).

See Also:

CategoricalResultFeature.getCategoryValues()