Filtrado colaborativo basado en deep learning para la recomendación de ítems en e-commerce

Abstract

Abstract: Recommender systems are very important nowadays, as they allow us to filter relevant information for the user, helping to improve the customer experience in e-commerce. One of the most widely used approaches to recommender systems is collaborative filtering, which relies on historical user information for recommendation. New ways to take better advantage of the amount of data available are currently being investigated, with the use of deep network architectures that are having a great impact in some areas of knowledge. These new architectures are being used to improve recommender systems, accurately reflecting user preferences. This work focuses its study on the collaborative filtering method based on deep neural networks, where the types of networks such as Autoencoder are analyzed for rating prediction, considering historical information from the user about how much he likes or dislikes an item.