org.lenskit.knn.user

Class UserUserItemScorer

java.lang.Object

org.lenskit.basic.AbstractItemScorer

org.lenskit.knn.user.UserUserItemScorer

All Implemented Interfaces:

ItemScorer

public class UserUserItemScorer
extends AbstractItemScorer

Score items with user-user collaborative filtering.

The detailed results returned by this scorer are of type UserUserResult.

Field Summary

Fields

Modifier and Type	Field and Description
protected NeighborFinder	neighborFinder
protected UserVectorNormalizer	normalizer

Constructor Summary

Constructors

Constructor and Description
UserUserItemScorer(RatingVectorPDAO rvd, NeighborFinder nf, UserVectorNormalizer norm, int nnbrs, int minNbrs, Threshold thresh)

Method Summary

Modifier and Type	Method and Description
protected Long2ObjectMap<? extends Collection<Neighbor>>	findNeighbors(long user, LongSet items) Find the neighbors for a user with respect to a collection of items.
protected Long2ObjectMap<SparseVector>	normalizeNeighborRatings(Collection<? extends Collection<Neighbor>> neighborhoods) Normalize all neighbor rating vectors, taking care to normalize each one only once.
ResultMap	scoreWithDetails(long user, Collection<Long> items) Score a collection of items and potentially return more details on the scores.

Methods inherited from class org.lenskit.basic.AbstractItemScorer

score, score

Methods inherited from class java.lang.Object

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

Field Detail

neighborFinder

protected final NeighborFinder neighborFinder

normalizer

protected final UserVectorNormalizer normalizer

Constructor Detail

UserUserItemScorer

@Inject
public UserUserItemScorer(RatingVectorPDAO rvd,
                                   NeighborFinder nf,
                                   UserVectorNormalizer norm,
                                   @NeighborhoodSize
                                   int nnbrs,
                                   @MinNeighbors
                                   int minNbrs,
                                   @UserSimilarityThreshold
                                   Threshold thresh)

Method Detail

normalizeNeighborRatings

protected Long2ObjectMap<SparseVector> normalizeNeighborRatings(Collection<? extends Collection<Neighbor>> neighborhoods)

Normalize all neighbor rating vectors, taking care to normalize each one only once.

FIXME: MDE does not like this method.

Parameters:

neighborhoods -

scoreWithDetails

@Nonnull
public ResultMap scoreWithDetails(long user,
                                            @Nonnull
                                            Collection<Long> items)

Description copied from interface: ItemScorer

Score a collection of items and potentially return more details on the scores.

Parameters:

user - The user ID for whom to generate scores.

items - The item to score.

Returns:

The scores for the items. This result set may not contain all requested items. Implementations that support additional details will return a subclass of ResultMap that provides access to those details.

findNeighbors

protected Long2ObjectMap<? extends Collection<Neighbor>> findNeighbors(long user,
                                                                       @Nonnull
                                                                       LongSet items)

Find the neighbors for a user with respect to a collection of items. For each item, the neighborhoodSize users closest to the provided user are returned.

Parameters:

user - The user’s rating vector.

items - The items for which neighborhoods are requested.

Returns:

A mapping of item IDs to neighborhoods.