Forecasting POI Occupation with Contextual Machine Learning

The increasing availability of historical information has emphasized the importance to explore, understand and extract value from it in order to achieve both short-term goals and strategic objectives. Intelligent techniques to handle heterogeneous data, together with user preferences, may be beneficial for end users; among them we can mention recommendation systems, which are able to guide users through huge catalogues of alternative items. This kind of systems represent an invaluable help not only for the users, who can feel disoriented in presence of so many alternatives at their disposal, but also for service providers or sellers, which can benefit from inferred hidden knowledge and guide towards particular items the choices of specific groups of users sharing some common preferences. This influence capability of recommendation systems can be particularly useful in the touristic domain, where the need to control and manage the level of crowding of POIs (Points Of Interest) has become a pressing need in the recent years. In this paper we study the role of contextual information in determining POI occupations and we explore how machine learning and deep learning technologies can help in producing good POI occupation forecasters by enriching historical information with contextual one. Throughout the paper we refer to a real-world application scenario regarding the touristic visits performed in Verona, a municipality in Northern Italy, between 2014 and 2019.