Having a Bad Day? Predicting High Delay Days in the National Airspace System

Experiencing high delays is a “bad day” for the National Airspace System (NAS). We apply machine learning algorithms to model the system delay and predict high delay days in the NAS for the 2010s. A broader scope of factors that may affect the system delay is examined, including queueing delays, terminal conditions, convective weather, wind, traffic volume, and special events. We train models to relate the system delay to these features spatially and temporally, and compare the performance of penalized regressions, kernelized support vector regressions, and ensemble regressions. The learned weights of the selected model reveal the spatial pattern and time consistency of the feature importance. Queuing delays, convective weather, and wind are found to be the most significant causative factors for system delays. We then identify high delay days using the model-predicted delay and observe an increasing trend over the past decade. The counterfactual analysis results suggest worsening convective weather after 2014, and a surge in demand in 2013 that was subsequently compensated by increased capacity.