recsys_metrics_polars.recall#

Classes

Recall@k

class recsys_metrics_polars.recall.RecallAtK(data_info)[source]#

Recall@k

$Recall@k = \dfrac{\text{Number of relevant items with rank} \leq k}{\min(k, \text{Total relevenat items for query})}$

fit(true_interactions, recommendations)[source]#

Prepare data for metric computing

Parameters:

true_interactions (DataFrame) – true interactions
recommendations (DataFrame) – predicted interactions with scores for each pair query and item

Return type:

BaseRecMetric

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

Compute metric per query

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

Compute mean metric value over all queries