6.2.Recommender Systems - Acknowledgement: Material derived...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Acknowledgement: Material derived from Adomavicius & Tuzhulin (2005) IEEE Trans. on Knowledge and Data Engineering, 17(6), 734–749. 11s1: COMP9417 Machine Learning and Data Mining Recommender Systems March 29, 2011 Aims Introduction This lecture will enable you to describe and reproduce machine learning approaches within the framework of Recommender Systems. Following it you should be able to: – “person who liked x may also like y ” • related to instance-based learning • define the problem of recommender systems • describe content-based, collaborative and hybrid recommender systems • reproduce key similarity-based approaches to recommender systems COMP9417: March 29, 2011 • Recommender systems – a form of personalization Recommender Systems: Slide 1 – similarity function • other forms of learning may be used to model user choices COMP9417: March 29, 2011 Recommender Systems: Slide 2 A Framework for Recommendation Content-based Recommendation User c is recommended items s that are similar to past choices. • idea comes from information retrieval • requires a profile of the content or description of items u(c, s) = score (ContentBasedProfile (c), Content (s)) Example movie rating matrix, where each entry has user c rating item s. E.g., Given: utility u : c × s ￿→ R u(c, s) = cos(wc, ws) = ￿￿ Problem: ∀c ∈ C , choose s￿ = argmaxs∈S u(c, s) c where wc · ws ￿￿ ￿wc￿2 × ￿ws￿2 ￿ ￿ This is learning in the sense of requiring extrapolation to predict the unknown values of the utility funcion. wc is a vector of summarising terms of c’s past choices, and ￿ COMP9417: March 29, 2011 COMP9417: March 29, 2011 Recommender Systems: Slide 3 ws is a vector of most relevant terms describing s ￿ Recommender Systems: Slide 4 Collaborative-based Recommendation Content-based Recommendation Advantages User c is recommended items that users with similar taste have chosen. • well-understood techniques from Information Retrieval • a.k.a. collaborative filtering (CF) • can extract latent features from text analysis • Amazon-style recommender systems Disadvantages • may not have content, or may be limited or sparse • over-specialisation: recommendations given for known types only • new user problem: must do some rating to get recommendations Two main methods: memory-based, and model-based CF. COMP9417: March 29, 2011 Recommender Systems: Slide 5 COMP9417: March 29, 2011 Recommender Systems: Slide 6 Memory-based CF Model-based CF Predict unknown rating rc,s of user c for item s by aggregating the ratings of N users c￿ most similar to c who have rated s: However, can use other machine learning methods to build a model to predict directly the unknown rating rc,s from examples in the database. rc,s = aggrc￿∈C rc￿,s E.g., Naive Bayes-type approaches (see later lecture.) What aggregation to use ? One commonly used is weighted sum rc,s = k ￿ c￿ ∈ C Memory-based CF is like a nearest-neighbour method. This is called model-based CF. ￿ sim(c, c ) × rc￿,s where k is just a normalising factor, and the similarity function can be correlation, cosine distance, etc. on the vector of items rated (e.g., bought) by users. Alternatively, can use item-based similarity (Amazon). COMP9417: March 29, 2011 Recommender Systems: Slide 7 COMP9417: March 29, 2011 Recommender Systems: Slide 8 Summary Collaborative-based Recommendation Advantages • based on techniques from information retrieval and machine learning • works well in practice • an application area growing rapidly • does not require content (descriptions) • simple systems can do surprisingly well • many possible extensions, e.g., recommendation in social networks Disadvantages • new user problem: must do some rating to get recommendations • new item problem: must be rated to be used in recommendations COMP9417: March 29, 2011 Recommender Systems: Slide 9 COMP9417: March 29, 2011 Recommender Systems: Slide 10 ...
View Full Document

This note was uploaded on 06/20/2011 for the course COMP 9417 taught by Professor Some during the Three '11 term at University of New South Wales.

Ask a homework question - tutors are online