com.graphaware.reco.generic.engine.DelegatingRecommendationEngine Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of recommendation-engine Show documentation
Show all versions of recommendation-engine Show documentation
Skeleton for a Recommendation Engine
The newest version!
/*
* Copyright (c) 2013-2016 GraphAware
*
* This file is part of the GraphAware Framework.
*
* GraphAware Framework is free software: you can redistribute it and/or modify it under the terms of
* the GNU General Public License as published by the Free Software Foundation, either
* version 3 of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
* without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
* See the GNU General Public License for more details. You should have received a copy of
* the GNU General Public License along with this program. If not, see
* extends BaseRecommendationEngine {
private final List> engines = new LinkedList<>();
private final List> postProcessors = new LinkedList<>();
public DelegatingRecommendationEngine() {
addEngines(engines());
addPostProcessors(postProcessors());
}
/**
* {@inheritDoc}
*/
@Override
public String name() {
return "Delegating Engine @" + this.hashCode();
}
/**
* Get {@link com.graphaware.reco.generic.engine.RecommendationEngine}s to be delegated to. Designed to be overridden.
*
* @return empty list by default.
*/
protected List> engines() {
return Collections.emptyList();
}
/**
* Get {@link com.graphaware.reco.generic.post.PostProcessor}s to be used by this engine. Designed to be overridden.
*
* @return empty list by default.
*/
protected List> postProcessors() {
return Collections.emptyList();
}
/**
* Get {@link com.graphaware.reco.generic.filter.BlacklistBuilder}s to be used by this factory. Designed to be overridden.
*
* @return empty list by default.
*/
protected List> blacklistBuilders() {
return Collections.emptyList();
}
/**
* Get {@link com.graphaware.reco.generic.filter.Filter}s to be used by this factory. Designed to be overridden.
*
* @return empty list by default.
*/
protected List> filters() {
return Collections.emptyList();
}
/**
* Add a {@link com.graphaware.reco.generic.engine.RecommendationEngine} that this engine delegates to. Delegation
* happens in the order in which engines are added.
*
* @param engine to delegate to. Must not be null.
*/
public final void addEngine(RecommendationEngine engine) {
notNull(engine);
engines.add(engine);
}
/**
* Add {@link com.graphaware.reco.generic.engine.RecommendationEngine}s that this engine delegates to, in the order
* in which they are added.
*
* @param engines to delegate to. Must not be null and all of the elements must not be null.
*/
public final void addEngines(List> engines) {
notNull(engines);
for (RecommendationEngine engine : engines) {
addEngine(engine);
}
}
/**
* Add a {@link com.graphaware.reco.generic.post.PostProcessor}s that is used to post-process recommendations once
* computed. The post-processors are applied in the order in which they are added.
*
* @param postProcessor to be used. Must not be null.
*/
public final void addPostProcessor(PostProcessor postProcessor) {
notNull(postProcessor);
postProcessors.add(postProcessor);
}
/**
* Add {@link com.graphaware.reco.generic.post.PostProcessor}s that are used to post-process recommendations once
* computed. The post-processors are applied in the order in which they are added.
*
* @param postProcessors to be used. Must not be null and all of the elements must not be null.
*/
public final void addPostProcessors(List> postProcessors) {
notNull(postProcessors);
for (PostProcessor postProcessor : postProcessors) {
addPostProcessor(postProcessor);
}
}
/**
* {@inheritDoc}
*/
@Override
public Recommendations doRecommend(IN input, Context context) {
Recommendations recommendations = new Recommendations<>();
for (RecommendationEngine engine : engines) {
if (engine.participationPolicy(context).participate(input, context, recommendations)) {
recommendations.merge(engine.recommend(input, context));
}
}
removeIrrelevant(input, context, recommendations);
for (PostProcessor postProcessor : postProcessors) {
postProcessor.postProcess(recommendations, input, context);
}
return recommendations;
}
/**
* Remove recommendations that have no chance of making it to the final selection, because their score will always
* be lower than the score of the last returned recommendation, even after post processing. This is a performance
* optimisation, so that irrelevant recommendations don't have to be post-processed.
*
* @param input input to the recommendation engine. Typically the person or item recommendations are being
* computed for.
* @param context additional information about the recommendation process.
* @param recommendations computed so far.
*/
private void removeIrrelevant(IN input, Context context, Recommendations recommendations) {
float maxRelativeChange = maxRelativeChange(input, context);
if (Float.POSITIVE_INFINITY == maxRelativeChange) {
return;
}
int i = 0;
float minScoreInLimit = 0;
for (Recommendation recommendation : recommendations.get(Integer.MAX_VALUE)) {
if (++i == context.config().limit()) {
minScoreInLimit = recommendation.getScore().getTotalScore() - maxRelativeChange;
} else if (i > context.config().limit() && recommendation.getScore().getTotalScore() < minScoreInLimit) {
recommendations.remove(recommendation.getItem());
}
}
}
/**
* Get the maximum value by which score difference between any two recommendations can be changed by post processing.
*
* @param input input to the recommendation engine. Typically the person or item recommendations are being
* computed for.
* @param context additional information about the recommendation process.
* @return maximum relative change. If unknown, {@link Float#POSITIVE_INFINITY} will be returned.
*/
private float maxRelativeChange(IN input, Context context) {
float result = 0f;
for (PostProcessor postProcessor : postProcessors) {
float posInfluence = postProcessor.maxPositiveScore(input, context);
float negInfluence = postProcessor.maxNegativeScore(input, context);
if (posInfluence < 0) {
throw new IllegalStateException(postProcessor + " has a negative influence score (" + posInfluence + "), should not be negative");
}
if (negInfluence > 0) {
throw new IllegalStateException(postProcessor + " has a positive influence score (" + negInfluence + "), should not be positive");
}
if (Float.isInfinite(negInfluence) || Float.isInfinite(posInfluence)) {
result = Float.POSITIVE_INFINITY;
break;
}
result += Math.abs(negInfluence) + Math.abs(posInfluence);
}
return result;
}
/**
* {@inheritDoc}
*/
@Override
public boolean equals(Object o) {
if (this == o) return true;
if (o == null || getClass() != o.getClass()) return false;
DelegatingRecommendationEngine that = (DelegatingRecommendationEngine) o;
if (!engines.equals(that.engines)) return false;
if (!postProcessors.equals(that.postProcessors)) return false;
return true;
}
/**
* {@inheritDoc}
*/
@Override
public int hashCode() {
int result = engines.hashCode();
result = 31 * result + postProcessors.hashCode();
return result;
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy