org.lenskit.predict.ordrec

Class OrdRecRatingPredictor

java.lang.Object

org.lenskit.basic.AbstractRatingPredictor

org.lenskit.predict.ordrec.OrdRecRatingPredictor

All Implemented Interfaces:

RatingPredictor

public class OrdRecRatingPredictor
extends AbstractRatingPredictor

OrdRec implementation of rating prediction.

The model views user feedback as ordinal. The framework is based on a pointwise (rather than pairwise) ordinal approach, it can wrap existing CF methods, and upgrade them into being able to tackle ordinal feedback. The implementation is based on Koren’s paper:

Since:

2.1

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	OrdRecRatingPredictor.FullResult The result type of OrdRec rating predictions.

Constructor Summary

Constructors

Constructor and Description
OrdRecRatingPredictor(ItemScorer scorer, DataAccessObject dao, Quantizer quantizer, double rate, double reg, int niters) Construct a new OrdRec rating predictor.

Method Summary

Modifier and Type	Method and Description
java.util.Map<java.lang.Long,java.lang.Double>	predict(long user, java.util.Collection<java.lang.Long> items) Predict the user’s preference for a collection of items.
ResultMap	predictWithDetails(long user, java.util.Collection<java.lang.Long> items) Compute detailed predictions for the user.

Methods inherited from class org.lenskit.basic.AbstractRatingPredictor

predict

Methods inherited from class java.lang.Object

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

Constructor Detail

OrdRecRatingPredictor

@Inject
public OrdRecRatingPredictor(ItemScorer scorer,
                                     DataAccessObject dao,
                                     Quantizer quantizer,
                                     @LearningRate
                                     double rate,
                                     @RegularizationTerm
                                     double reg,
                                     @IterationCount
                                     int niters)

Construct a new OrdRec rating predictor.

Parameters:

scorer - The ItemScorer to produce the underlyign scores.

dao - The DAO to access user events.

quantizer - The quantizer to which ratings should be constrained.

rate - The learning rate for user profile training.

reg - Regularization term for user profile training.

Method Detail

predict

@Nonnull
public java.util.Map<java.lang.Long,java.lang.Double> predict(long user,
                                                                       @Nonnull
                                                                       java.util.Collection<java.lang.Long> items)

Description copied from class: AbstractRatingPredictor

Predict the user’s preference for a collection of items.

This implementation delegates to RatingPredictor.predictWithDetails(long, Collection).

Specified by:

predict in interface RatingPredictor

Overrides:

predict in class AbstractRatingPredictor

Parameters:

user - The user ID for whom to generate predicts.

items - The items to predict for.

Returns:

A mapping from item IDs to predicted preference. This mapping may not contain all requested items.

predictWithDetails

@Nonnull
public ResultMap predictWithDetails(long user,
                                             @Nonnull
                                             java.util.Collection<java.lang.Long> items)

Compute detailed predictions for the user.

Parameters:

user - The user ID for whom to generate predictions.

items - The items to predict for.

Returns:

The detailed results; each result is an instance of OrdRecRatingPredictor.FullResult.