Recommender system

Software that suggests items (movies, products, posts) to users from past behavior and similarity patterns, often using collaborative filtering or matrix factorization.

A recommender system predicts what a user might like next: which movie to stream, which product to buy, or which article to read. Many industrial systems rely on collaborative filtering, inferring taste from ratings or clicks shared across users and items.

The Netflix Prize (2006 to 2009) turned recommender accuracy into a public benchmark measured with RMSE. Winning teams combined matrix factorization, ensemble methods, and restricted Boltzmann machines from Hinton’s group. Recommenders remain a core supervised learning application in machine learning.