Recommender Systems and Techniques

Recommender systems are the computational methods that decide what a user sees next — which product appears in a search result, which video plays after the current one, which article a news feed surfaces first. Researchers in this area study how to infer what a person is likely to want by combining signals from their past behavior, the behavior of similar users, and the properties of the items themselves, using techniques ranging from classical matrix factorization to modern deep neural networks. A persistent challenge is the cold-start problem: making useful recommendations for new users or new items when little behavioral data yet exists. Active research directions include building systems that adapt to the context in which a recommendation is made — time of day, location, current intent — and developing approaches that preserve user privacy while still learning accurate preference models.

Works

73,925

Total citations

1,096,353

Keywords

Collaborative FilteringMatrix FactorizationDeep LearningContent-Based RecommendationWeb MiningContext-Aware Recommender Systems