org.lenskit.eval.traintest.recommend

Class TopNPrecisionRecallMetric

java.lang.Object

org.lenskit.eval.traintest.metrics.Metric<X>

org.lenskit.eval.traintest.recommend.TopNMetric<X>

org.lenskit.eval.traintest.recommend.ListOnlyTopNMetric<TopNPrecisionRecallMetric.Context>

org.lenskit.eval.traintest.recommend.TopNPrecisionRecallMetric

public class TopNPrecisionRecallMetric
extends ListOnlyTopNMetric<TopNPrecisionRecallMetric.Context>

A metric to compute the precision and recall of a recommender given a set of candidate items to recommend from and a set of desired items. The aggregate results are means of the user results.

This can be used to compute metrics like fallout (probability that a recommendation is bad) by configuring bad items as the test item set.

This metric is registered under the name pr. It has two configuration parameters:

suffix

a suffix to append to the column name

goodItems

an item selector expression. The default is the user’s test items.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	TopNPrecisionRecallMetric.Context
static class	TopNPrecisionRecallMetric.PresRecResult

Constructor Summary

Constructors

Constructor and Description
TopNPrecisionRecallMetric() Construct a new precision-recall metric using the user’s test items as good.
TopNPrecisionRecallMetric(ItemSelector good, java.lang.String sfx) Construct a new recall and precision top n metric
TopNPrecisionRecallMetric(PRMetricSpec spec) Construct a precision-reacll metric from a spec.

Method Summary

Modifier and Type	Method and Description
TopNPrecisionRecallMetric.Context	createContext(AlgorithmInstance algorithm, DataSet dataSet, RecommenderEngine rec) Create the context for an experimental condition (algorithm/data set pair).
MetricResult	getAggregateMeasurements(TopNPrecisionRecallMetric.Context context) Get the aggregate results from an accumulator.
MetricResult	measureUser(Recommender rec, TestUser user, int targetLength, LongList recs, TopNPrecisionRecallMetric.Context context) Measurement method that only uses the recommend list.

Methods inherited from class org.lenskit.eval.traintest.recommend.ListOnlyTopNMetric

measureUser

Methods inherited from class org.lenskit.eval.traintest.metrics.Metric

getAggregateColumnLabels, getColumnLabels, getRequiredRoots

Methods inherited from class java.lang.Object

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

Constructor Detail

TopNPrecisionRecallMetric

public TopNPrecisionRecallMetric()

Construct a new precision-recall metric using the user’s test items as good.

TopNPrecisionRecallMetric

public TopNPrecisionRecallMetric(PRMetricSpec spec)

Construct a precision-reacll metric from a spec.

Parameters:

spec - The precision-recall metric.

TopNPrecisionRecallMetric

public TopNPrecisionRecallMetric(ItemSelector good,
                                 java.lang.String sfx)

Construct a new recall and precision top n metric

Parameters:

good - an item selector for the good items.

sfx - the suffix label for this evaluation, or null for no suffix.

Method Detail

measureUser

@Nonnull
public MetricResult measureUser(Recommender rec,
                                         TestUser user,
                                         int targetLength,
                                         LongList recs,
                                         TopNPrecisionRecallMetric.Context context)

Description copied from class: ListOnlyTopNMetric

Measurement method that only uses the recommend list.

Thread Safety: This method may be called concurrently by multiple threads with the same recommender and context.

Specified by:

measureUser in class ListOnlyTopNMetric<TopNPrecisionRecallMetric.Context>

Parameters:

rec - The recommender used to recommend for this user.

user - The user.

targetLength - The target list length.

recs - The list of recommendations.

context - The context.

Returns:

The results of measuring this user.

createContext

@Nullable
public TopNPrecisionRecallMetric.Context createContext(AlgorithmInstance algorithm,
                                                                 DataSet dataSet,
                                                                 RecommenderEngine rec)

Description copied from class: Metric

Create the context for an experimental condition (algorithm/data set pair). The default implementation returns null.

Note: Contexts must be thread-safe, in that multiple concurrent calls to the appropriate user-measurement function with the same context must be safe. This can be handled either by the context itself, or by the user-measurement function.

Overrides:

createContext in class Metric<TopNPrecisionRecallMetric.Context>

Parameters:

algorithm - The algorithm.

dataSet - The data set.

rec - The LensKit recommender engine, if applicable. This can be null for an external algorithm that does not provide a LensKit recommender.

Returns:

The context. This will be passed to the individual measurement methods. If the metric does need to accumulate any results, this method can return null.

getAggregateMeasurements

@Nonnull
public MetricResult getAggregateMeasurements(TopNPrecisionRecallMetric.Context context)

Description copied from class: Metric

Get the aggregate results from an accumulator. The default implementation returns MetricResult.empty().

Overrides:

getAggregateMeasurements in class Metric<TopNPrecisionRecallMetric.Context>

Parameters:

context - The context for an experimental condition.

Returns:

The aggregate results from the accumulator.