es.uam.eps.ir.ranksys.diversity.binom.metrics

Class BinomialMetric<U,I,F>

java.lang.Object

es.uam.eps.ir.ranksys.metrics.AbstractRecommendationMetric<U,I>

es.uam.eps.ir.ranksys.diversity.binom.metrics.BinomialMetric<U,I,F>

Type Parameters:

U - type of the users

F - type of the items

I - type of the features

All Implemented Interfaces:

RecommendationMetric<U,I>

Direct Known Subclasses:

BinomialCoverage, BinomialDiversity, BinomialNonRedundancy

public abstract class BinomialMetric<U,I,F>
extends AbstractRecommendationMetric<U,I>

Abstract class for metrics using the binomial model. S. Vargas, L. Baltrunas, A. Karatzoglou, P. Castells. Coverage, redundancy and size-awareness in genre diversity for Recommender Systems. RecSys 2014.

Author:

Saúl Vargas (saul.vargas@uam.es)

Field Summary

Fields

Modifier and Type	Field and Description
protected RelevanceModel<U,I>	relModel relevance model

Constructor Summary

Constructors

Constructor and Description
BinomialMetric(BinomialModel<U,I,F> binomialModel, FeatureData<I,F,?> featureData, int cutoff, RelevanceModel<U,I> relModel) Constructor.

Method Summary

Modifier and Type	Method and Description
double	evaluate(Recommendation<U,I> recommendation) Returns a score for the recommendation list.
protected abstract double	getResultFromCount(BinomialModel.UserBinomialModel prob, it.unimi.dsi.fastutil.objects.Object2IntMap<F> count, int nrel, int nret) Result of the metric based on the number of times each features appears in a recommendation list.

Methods inherited from class java.lang.Object

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

Field Detail

relModel

protected final RelevanceModel<U,I> relModel

relevance model

Constructor Detail

BinomialMetric

public BinomialMetric(BinomialModel<U,I,F> binomialModel,
                      FeatureData<I,F,?> featureData,
                      int cutoff,
                      RelevanceModel<U,I> relModel)

Constructor.

Parameters:

binomialModel - binomial diversity model

featureData - feature data

cutoff - maximum length of the recommendation list to be evaluated

relModel - relevance model

Method Detail

evaluate

public double evaluate(Recommendation<U,I> recommendation)

Returns a score for the recommendation list.

Parameters:

recommendation - recommendation list

Returns:

score of the metric to the recommendation

getResultFromCount

protected abstract double getResultFromCount(BinomialModel.UserBinomialModel prob,
                                             it.unimi.dsi.fastutil.objects.Object2IntMap<F> count,
                                             int nrel,
                                             int nret)

Result of the metric based on the number of times each features appears in a recommendation list.

Parameters:

prob - user binomial model

count - count map of each feature in a recommendation

nrel - number of relevant items in the recommendation

nret - length of the recommendation

Returns:

value of the metric