User–user collaborative filtering, also known as k-NN collaborative filtering,
was the first of the automated CF methods. It was first introduced
in the GroupLens Usenet article recommender [119]. The Ringo
music recommender [137] and the BellCore video recommender [62] also
used user-user CF or variants thereof.
User–user CF is a straightforward algorithmic interpretation of the
core premise of collaborative filtering: find other users whose past rating
behavior is similar to that of the current user and use their ratings
on other items to predict what the current user will like. To predict
Mary’s preference for an item she has not rated, user–user CF looks
for other users who have high agreement with Mary on the items they
have both rated. These users’ ratings for the item in question are then
weighted by their level of agreement with Mary’s ratings to predict
Mary’s preference.
Besides the rating matrix R, a user–user CF system requires a similarity
function s:U×U → R computing the similarity between two users
and a method for using similarities and ratings to generate predictions.