Collaborative Filtering via Concept Decomposition on the Netflix Dataset

Collaborative filtering recommender systems make automatic predictions about the interests of a user by collectin g information from many users (collaborating). Most recommendation algorithms are based in finding sets of customers or items whose ra tings overlap in order to create a model for inferring future ratin gs or items that might be of interest for a particular user. Traditional collaborative filtering techniques such as k-Nearest Neighbours and S i gular Value Decomposition (SVD) usually provide good accuracy bu t are computationally very expensive. The Netflix Prize is a colla borative filtering problem whose dataset is much larger than the previ ously known benchmark sets and thus traditional methods are stres sed to their limits when challenged with a dataset of that size. In t his paper we present experimental results that show how the concept decomposition method performs on the movie rating prediction task over the Netflix dataset and we show that it is able to achieve a well bal anced performance between computational complexity and predict ion accuracy.