org.grouplens.lenskit.basic

Class TopNItemRecommender

java.lang.Object

org.grouplens.lenskit.basic.AbstractItemRecommender

org.grouplens.lenskit.basic.TopNItemRecommender

All Implemented Interfaces:

ItemRecommender

Direct Known Subclasses:

ScoreBasedItemRecommender

public class TopNItemRecommender
extends AbstractItemRecommender

Recommender that recommends the top N items by a scorer. Implements all methods required by AbstractItemRecommender. The default exclude set is all items rated by the user.

Recommendations are returned in descending order of score.

Since:

1.1

Author:

GroupLens Research

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	TopNItemRecommender.Provider An intelligent provider for Top-N recommenders.

Field Summary

Fields

Modifier and Type	Field and Description
protected ItemScorer	scorer

Fields inherited from class org.grouplens.lenskit.basic.AbstractItemRecommender

dao

Constructor Summary

Constructors

Constructor and Description
TopNItemRecommender(DataAccessObject dao, ItemScorer scorer)

Method Summary

Methods

Modifier and Type	Method and Description
protected LongSet	getDefaultExcludes(long user) Get the default exclude set for a user.
protected LongSet	getDefaultExcludes(UserHistory<? extends Event> user) Get the default exclude set for a user.
protected LongSet	getPredictableItems(long user) Determine the items for which predictions can be made for a certain user.
ItemScorer	getScorer()
protected ScoredLongList	recommend(int n, SparseVector scores) Pick the top n items from a score vector.
protected ScoredLongList	recommend(long user, int n, LongSet candidates, LongSet exclude) Implement the ID-based recommendation in terms of the scorer.
protected ScoredLongList	recommend(UserHistory<? extends Event> user, int n, LongSet candidates, LongSet exclude) Implement profile-based recommendation in terms of the scorer.

Methods inherited from class org.grouplens.lenskit.basic.AbstractItemRecommender

canUseHistory, getUserHistory, recommend, recommend, recommend, recommend, recommend, recommend, recommend, recommend

Methods inherited from class java.lang.Object

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

Field Detail

scorer

protected final ItemScorer scorer

Constructor Detail

TopNItemRecommender

@Inject
public TopNItemRecommender(DataAccessObject dao,
                          ItemScorer scorer)

Method Detail

getScorer

public ItemScorer getScorer()

recommend

protected ScoredLongList recommend(long user,
                       int n,
                       LongSet candidates,
                       LongSet exclude)

Implement the ID-based recommendation in terms of the scorer. This method uses getDefaultExcludes(long) to supply a missing exclude set.

Overrides:

recommend in class AbstractItemRecommender

Parameters:

user - The user ID.

n - The number of items to return, or negative to return all possible items.

candidates - The candidate set.

exclude - The set of excluded items, or null to use the default exclude set.

Returns:

A scored list of item IDs.

See Also:

ItemRecommender.recommend(long, int, Set, Set)

recommend

protected ScoredLongList recommend(UserHistory<? extends Event> user,
                       int n,
                       LongSet candidates,
                       LongSet exclude)

Implement profile-based recommendation in terms of the scorer. This method uses getDefaultExcludes(long) to supply a missing exclude set.

Specified by:

recommend in class AbstractItemRecommender

Parameters:

user - The user rating vector.

n - The number of items to return, or negative to return all possible items.

candidates - The candidate set.

exclude - The set of excluded items, or null for the default exclude set.

Returns:

The recommendations with associated scores.

See Also:

ItemRecommender.recommend(UserHistory, int, Set, Set)

recommend

protected ScoredLongList recommend(int n,
                       SparseVector scores)

Pick the top n items from a score vector.

Parameters:

n - The number of items to recommend.

scores - The scored item vector.

Returns:

The top n items from scores, in descending order of score.

getDefaultExcludes

protected LongSet getDefaultExcludes(long user)

Get the default exclude set for a user. The base implementation gets all their rated items.

Parameters:

user - The user ID.

Returns:

The set of items to exclude.

getDefaultExcludes

protected LongSet getDefaultExcludes(UserHistory<? extends Event> user)

Get the default exclude set for a user. The base implementation returns all their rated items.

Parameters:

user - The user history.

Returns:

The set of items to exclude.

getPredictableItems

protected LongSet getPredictableItems(long user)

Determine the items for which predictions can be made for a certain user. This implementation is naive and asks the DAO for all items; subclasses should override it with something more efficient if practical.

Parameters:

user - The user's ID.

Returns:

All items for which predictions can be generated for the user.