Recommender Systems Introduction
Manuel Dömer
Abteilung Angewandte Mathematik, Physik und Operations
Items & Users
- Webshop
- Items: Products
- Users: Customers
- Library
- Items: Books
- Users: Library visitors
- News site
- Items: News articles
- Users: Readers
Recommender Systems
Suggest personalised and relevant items
- Cover the entire spectrum of the user’s interests
- Take the user’s context into account
- Avoid suggestions from what is already known
- Expand the user’s range of interests
Scalability of Recommender Systems
Aproaches for Recommender Systems
- Popularity-based
- Association rules mining
- Content-based
- Collaborative filtering
Popularity-Based recommendations
Example: The New York Times Online
Pros:
- Small complexity
- Explainable
Cons:
- No personalisation
- Item properties are not taken into account