Predicting Aircraft Trajectory Choice – A Nominal Route Approach

In this work, we propose a novel approach to predict aircraft trajectory choice. A trajectory clustering technique is used to consolidate historical flight tracks into a small set, and the cluster assignment results are then used as the ground truth of the route choice. Three types of features are used to predict the trajectory choice: convective weather, wind, and Miles-In-Trail (MIT) restrictions. Dimension of the features is greatly reduced by matching them with the representative trajectories of different clusters, which we call Nominal Routes. Four popular machine learning models are explored and compared: logistic regression, support vector machine, random forest, and gradient boosting. We apply our methods to five airport pairs: IAH to BOS, BOS to IAH, FLL to JFK, JFK to FLL, and LAX to SEA. The random forest approach has the best performance for all pairs except IAH to BOS, where gradient boosting has slightly better performance. Based on the best models, we rank the importance of features for different airport pairs. Results vary, but in general, wind has the largest effect, followed by thunderstorm, rain, and MIT.