lenskit.metrics#

Metrics for evaluating recommender outputs.

Submodules#

`basic`	Basic set statistics.
`bulk`
`predict`	Prediction accuracy metrics. See eval-predict-accuracy for an overview
`ranking`	LensKit ranking (and list) metrics.
`reranking`	LensKit reranking metrics.

Attributes#

`MetricResult`	Results of a metric, either a single value or a dictionary of values.
`MetricVal`	A single metric value.

Classes#

`ListMetric`	Base class for metrics defined on individual recommendation outputs. This
`Metric`	Base class for LensKit metrics. Simple metrics that compute a single value
`MetricFunction`	Interface for per-list metrics implemented as simple functions.
`MeasurementCollector`	Collect metric measurements over multiple recommendation lists.
`RunMetrics`	Results of measuring a single run in a metric collector.

Functions#

quick_measure_model(model, data, *[, ...])

Do a quick-and-dirty model measurement with a default pipeline setup, split,

Package Contents#

type lenskit.metrics.MetricResult = MetricVal | Mapping[str, MetricVal]#: Results of a metric, either a single value or a dictionary of values.

type lenskit.metrics.MetricVal = float | int | object#: A single metric value.

lenskit.metrics.quick_measure_model(model, data, *, predicts_ratings=False, n_jobs=1, rng=None)#

Do a quick-and-dirty model measurement with a default pipeline setup, split, and metrics. This is mostly to make tests easy to write, you usually don’t want to use it for actual recommender evaluation.

Stability:

Caller (see Stability Levels).

Parameters:

model (lenskit.pipeline.Component)
data (lenskit.data.Dataset)
predicts_ratings (bool)
n_jobs (int | None)
rng (lenskit.random.RNGInput)

Return type:

lenskit.metrics.bulk.RunAnalysisResult

Exported Aliases#

class lenskit.metrics.ItemList#: Re-exported alias for lenskit.data.ItemList.

class lenskit.metrics.ListLength#: Re-exported alias for lenskit.metrics.basic.ListLength.

class lenskit.metrics.TestItemCount#: Re-exported alias for lenskit.metrics.basic.TestItemCount.

class lenskit.metrics.RunAnalysis#: Re-exported alias for lenskit.metrics.bulk.RunAnalysis.

class lenskit.metrics.RunAnalysisResult#: Re-exported alias for lenskit.metrics.bulk.RunAnalysisResult.

class lenskit.metrics.MAE#: Re-exported alias for lenskit.metrics.predict.MAE.

class lenskit.metrics.RMSE#: Re-exported alias for lenskit.metrics.predict.RMSE.

class lenskit.metrics.DCG#: Re-exported alias for lenskit.metrics.ranking.DCG.

class lenskit.metrics.ILS#: Re-exported alias for lenskit.metrics.ranking.ILS.

class lenskit.metrics.NDCG#: Re-exported alias for lenskit.metrics.ranking.NDCG.

class lenskit.metrics.RBP#: Re-exported alias for lenskit.metrics.ranking.RBP.

class lenskit.metrics.Entropy#: Re-exported alias for lenskit.metrics.ranking.Entropy.

class lenskit.metrics.ExposureGini#: Re-exported alias for lenskit.metrics.ranking.ExposureGini.

class lenskit.metrics.GeometricRankWeight#: Re-exported alias for lenskit.metrics.ranking.GeometricRankWeight.

class lenskit.metrics.Hit#: Re-exported alias for lenskit.metrics.ranking.Hit.

class lenskit.metrics.ListGini#: Re-exported alias for lenskit.metrics.ranking.ListGini.

class lenskit.metrics.LogRankWeight#: Re-exported alias for lenskit.metrics.ranking.LogRankWeight.

class lenskit.metrics.MeanPopRank#: Re-exported alias for lenskit.metrics.ranking.MeanPopRank.

class lenskit.metrics.Precision#: Re-exported alias for lenskit.metrics.ranking.Precision.

class lenskit.metrics.RankBiasedEntropy#: Re-exported alias for lenskit.metrics.ranking.RankBiasedEntropy.

class lenskit.metrics.RankingMetricBase#: Re-exported alias for lenskit.metrics.ranking.RankingMetricBase.

class lenskit.metrics.RankWeight#: Re-exported alias for lenskit.metrics.ranking.RankWeight.

class lenskit.metrics.Recall#: Re-exported alias for lenskit.metrics.ranking.Recall.

class lenskit.metrics.RecipRank#: Re-exported alias for lenskit.metrics.ranking.RecipRank.

lenskit.metrics.least_item_promoted()#: Re-exported alias for lenskit.metrics.reranking.least_item_promoted().

lenskit.metrics.rank_biased_overlap()#: Re-exported alias for lenskit.metrics.reranking.rank_biased_overlap().