Overview of methods of collaborative filtering

Article can be useful for researchers in such areas as big data and creation of recommender systems. Correlation and latent models based methods are covered. For correlation model, user-based and item-based methods are compared by conduction of analysis of their advantages and disadvantages. Latent model is explained as an opposite to correlation model, which is free of the disadvantages of the latter. Method of alternating least squares is explored. Described methods are tested with frameworks GraphLab and Apache Spark on open dataset MovieLens. Stated problem is to estimate film rate by user.