The Comparison Study of Matrix Factorization on Collaborative Filtering Recommender System

The recommendation system has been a vital study topic in recent years, so many scientists and academics across the world are interested in researching the subject. Music, movies, books, news, commercial items, and search inquiries are all examples of applying the recommendation system. One of the most common and successful strategies in recommendation systems is collaborative filtering. This method seeks to find similar users who are active to recommend an item. The author proposes to compare model-based collaborative filtering techniques using matrix factorization algorithms, such as alternating least squares (ALS), singular value decomposition (SVD), Alternating Least Squares weight regularization (ALS-WR) and SVD++. Furthermore, the quality of the recommendation system is determined using three datasets with various features. According to the research findings of four algorithms (ALS, ALS-WR, SVD, SVD++), only the MovieLens dataset can overcome the sparsity problem, as seen by the RMSE score below 1.