Recommendation systems help users discover content and products that closely fit their specific preferences and interests, thus making it easier for them to find relevant information in large volumes of information. Nowadays, researchers are working with deep network architectures to offer recommendations, which can significantly impact some sectors. This paper presents a comparative analysis of the performance of Machine Learning and Deep Learning models in recommender systems, focusing on traditional techniques such as Matrix Factorization and advanced models such as Autoencoder and Neural collaborative filtering. Experiments carried out on two datasets demonstrate, that although deep learning models offer a higher prediction quality using root mean square error, their performance in terms of precision and recall varies depending on the dataset and the number of recommendations generated.