Revisión y análisis de sistemas recomendadores basados en sesión

Abstract

[en] Due to the wide range of choice, humans have always relied on experts for their selection. The recommendations that could be made physically in a shop or centre have been replaced by automatic systems. They recognise the user's tastes and, based on the information gathered, try to predict good recommendations for the user. In the last few years there has been a revolution in this matter due to improvements in mainly Machine learning, therefore there is a need to categorise and evaluate the new contributions to the state of the art to check if these new methods offer an improvement to the traditional ones. The main focus of the work is to recognise, interpret and categorize these new systems that predict through the information collected during a session or also known as Session Based Recommenders, specifically we will focus on those based on Deep Learning. As a result of the project, a review of the current state of the art and an in-depth study of four novel methods is proposed, with a theoretical and practical analysis of the results. It is thanks to this study and all the publications that support it that it can be affirmed that there is a revolution in this subject and that all the methods studied in this work offer an improvement on the traditional approach.