Real-time prediction of flight arrival times using surveillance information

Accurate estimations of the time of arrival of a flight prior to its landing are beneficial for most of the stakeholders in the air traffic management and control industry, as they could lead to reductions in potential safety risks and improvements in resources allocation. In this paper, and within the European Commission funded Transforming Transport project, we propose a methodology for real-time prediction of flight arrival times based on the application of machine learning techniques. For this purpose, we employ state-of-the-art data warehousing and broadcasting processes, that allow both the training of a regression machine learning model and the integration of its predictions of current flights on a real-time visualization tool set up for customer usage. The model only makes use of the information included in aircraft surveillance messages. Predictions obtained with such model are compared to those provided by other current services to observe the added value of the application of the proposed system on real-time operations.