lenskit.metrics.reranking

LensKit reranking metrics.

Functions

least_item_promoted(reference, reranked[, n])	Compute the Least Item Promoted (LIP) metric.
rank_biased_overlap(reference, reranked[, weight, n])	Computes the Rank Biased Overlap (RBO) between two item rankings.

Package Contents

lenskit.metrics.reranking.least_item_promoted(reference, reranked, n=10)

Compute the Least Item Promoted (LIP) metric.

This metric identifies the item in the top-k reranked list that had the highest (worst) rank in the reference, and returns how many positions it was promoted from beyond k.

Parameters:

• reference (lenskit.data.ItemList) – The original/base ranking.

• reranked (lenskit.data.ItemList) – The reranked list.

• n (int) – The depth to evaluate (default 10).

Returns:

The rank distance of the least-promoted item, or NaN if the base ranking is empty.

Stability:

Experimental

Return type:

float

lenskit.metrics.reranking.rank_biased_overlap(reference, reranked, weight=None, n=10)

Computes the Rank Biased Overlap (RBO) between two item rankings. RBO is defined by Webber et al. [WMZ10].

Parameters:

• reference (lenskit.data.ItemList) – The first item list to compare.

• reranked (lenskit.data.ItemList) – The second item list to compare.

• weight (lenskit.metrics.RankWeight | None) – The rank weighting to use. If None, defaults to GeometricRankWeight(0.85).

• n (int) – The depth to which to compute the overlap (default 10).

Returns:

The RBO score between 0 and 1.

Stability:

Experimental

Return type:

float