SFRUTTARE LE INFORMAZIONI SUL TEMPO E LE INFORMAZIONI SUI CONTENUTI PER MIGLIORARE LE RACCOMANDAZIONI BASATE SU APPROCCI COLLABORATIVI E DELLA CONOSCENZA

In the era of big data and artificial intelligence (i.e., AI), the massive data availability has created a tough world to navigate for users. Fortunately, big data is also the fuel of machine learning applications and modern powerful tools like recommender systems which automatically guides users towards options that align with their preferences. Furthermore, these systems can be applied to various domains, providing not only quietness to users during their decision process but also powerful tools for companies to understand and enrich their services and content. This was made possible by the intrinsic and hidden information structure that connects the data. It is thanks to this new understanding of data that many sectors, ranging from industry and transport to healthcare, have begun to preserve the knowledge derived from it. However, this data knowledge requires more optimized and faster methods to extract value, which is why new systems based on deep learning have emerged. This thesis first seeks to establish a solid foundation of knowledge to address state of the art recommendation systems more easily. In addition, we propose our own model that builds its structure on the temporal relationship that each user has with the content they interact with, seeking to extract new links and hidden information between these contents. All our findings are supported by experiments and analyses on a wide range of recommendation systems. Finally, this dissertation is enriched by an extensive literature review on recommender systems, which not only allows the reader to understand their potential, but also provides a valuable tool for the development of new models based on these strategies.