A Machine Learning Approach to Air Traffic Route Choice Modelling

Air Traffic Flow and Capacity Management (ATFCM) is one of the constituent parts of Air Traffic Management (ATM). The goal of ATFCM is to make airport and airspace capacity meet traffic demand and, when capacity opportunities are exhausted, optimise traffic flows to meet the available capacity. One of the key enablers of ATFCM is the accurate estimation of future traffic demand. The available information (schedules, flight plans, etc.) and its associated level of uncertainty differ across the different ATFCM planning phases, leading to qualitative differences between the types of forecasting that are feasible at each time horizon. While abundant research has been conducted on tactical trajectory prediction (i.e., during the day of operations), trajectory prediction in the pre-tactical phase, when few or no flight plans are available, has received much less attention. As a consequence, the methods currently in use for pre-tactical traffic forecast are still rather rudimentary, which often results in suboptimal ATFCM decision making. This paper proposes a machine learning approach for the prediction of airlines route choices between two airports as a function of the characteristics of each route, such as flight efficiency, air navigation charges and expected level of congestion. Different predictive models based on multinomial logistic regression and decision trees are formulated and calibrated using historical traffic data, and a critical evaluation of each model is conducted. For this purpose, we analyse the predictive power of each model in terms of its ability to forecast traffic volumes at the level of charging zones, showing that the proposed approach entails significant potential to enhance pre-tactical traffic forecast. We conclude by discussing the limitations and room for improvement of the proposed approach, as well as the future developments required to produce reliable traffic forecasts at a higher spatial and temporal resolution.