Collaborative Filtering Recommendation Systems Based on Deep Learning: An Experimental Study

Recommender systems allow users to filter relevant information, helping users discover content and products that fit their preferences and interests. Collaborative filtering is one of the most widely used approaches in recommender systems, which uses historical user data for the recommendation. Nowadays, researchers are exploring new ways to make recommendations, using deep network architectures that have a major impact on some areas. This paper explores collaborative filtering recommendation systems based on deep learning, focusing on the experimental evaluation of algorithms most used in these systems. In our experimental study, we evaluate the performance of Autoencoder and Neural Collaborative Filtering models on representative datasets. As the main result, we found that the Autoencoder model outperformed Neural Collaborative Filtering in terms of prediction, suggesting its usefulness by providing more precise recommendations for users.