The growing need for personalized information has resulted in the development of recommender systems (RS), which are widely used to predict user preferences and generate appropriate recommendations through the adoption of different techniques and algorithms. These systems can achieve the recommendation process using techniques based on Artificial Intelligence (AI). It can determine what users prefer based on their preferences and past behavior. Our work aims to make a comparative study of three matrix factorization (MF) methods, namely, singular value decomposition (SVD), singular value decomposition plus plus (SVD++) and non-negative matrix factorization (NMF), to identify the most efficient algorithm in terms of prediction in a movie recommender system based on collaborative filtering. The performance of these methods is experimentally verified over the MovieLens 100 K dataset by using the mean absolute error (MAE) and root mean square error (RMSE). The results showed the efficiency of SVD++ with a reduced values of MAE and RMSE.