Recall at K

Also known as · Recall@K

The share of all relevant items that appear in the top K recommendations: relevant items in the top K divided by total relevant items.

When to use

Asks how many good items the recommender missed, so it suits search and discovery, whereas Precision at K suits a homepage. The lecture notes that improving one often hurts the other, echoing Precision and Recall from classification.

See DS Lec 6.

Appears in

No references yet.