Implementing collaborative filtering (https://en.wikipedia.org/wiki/Collaborative_filtering) over ratings data set. By given user id and item id we estimate the probability for the user to like or not that item. Estimating the error by produced results:

Used Euclidian distance and Pearson correlativity for estimating the similarity.

Used libraries: NReco Recommender - .NET port of Apache Mahout CF java engine.