recsys_metrics_polars.avg_precision#

Classes

AvgPrecisionAtK(data_info)

Average precision and MAP (see AvgPrecisionAtK.avergae_over_queries())

class recsys_metrics_polars.avg_precision.AvgPrecisionAtK(data_info)[source]#

Bases: BaseMetricAtK

Average precision and MAP (see AvgPrecisionAtK.avergae_over_queries())

AP@k

$& AP@k = \dfrac{\sum\limits_{i=1}^{k} Precision@i \cdot Rel(i)}{\min(k, \text{Total relevant items})}, \\ & Rel(i) = \begin{cases} 1, \text{item with rank } i \text{ is relevant,} \\ 0, \text{otherwise} \end{cases}$

Parameters:: data_info (DataInfo) –

compute_per_query(k, **kwargs)[source]#

Compute metric per query

Return type:: DataFrame
Parameters:: k (int) –

avergae_over_queries(k, **kwargs)[source]#

Compute MAP@k :rtype: float

$MAP@k = \dfrac{\sum\limits_{i=1}^N AP_i@k}{N},$

$AP_i@k$ - average precision for query $i$ , $i \in {1,2,\ldots,N}, N-$ total number of queries.

Parameters:: k (int) –
Return type:: float