Signal processing and machine learning for air traffic delay predition

• Víctor M. Tenorio ^[1] ; Antonio G. Marques ^[1] ; Luis Cadarso ^[1]
  1. [1] Universidad Rey Juan Carlos

     Universidad Rey Juan Carlos

     Madrid, España

     
• Localización: R-evolucionando el transporte [Recurso electrónico]: XIV Congreso de Ingeniería del Transporte. Universidad de Burgos 6, 7 y 8 de julio 2021 / coord. por Hernán Gonzalo Orden, Marta Rojo Arce, 2021, ISBN 978-84-18465-12-3, págs. 2603-2614
• Idioma: inglés
• Enlaces
  • Texto Completo Libro
• Resumen

  • As data quality and quantity increase, the prediction of future events using machine learning (ML) techniques across engineering disciplines grows by the day. Air transportation cannotbe an exception. Delay prediction is paramount in the aerospace industry, since air trafficdelays are responsible for millions of dollars in losses to airlines and passengers, along withnegative impacts on the environment. In this contribution, we leverage recent signalprocessing and ML advances to put forth a processing-and-learning pipeline for theprediction of air traffic delays. The proposed approach is executed in several steps. Firstly,we apply signal processing and data science techniques to filter and denoise the originalinformation. Secondly, we run a descriptive analysis of the data and design new featurestailored to the prediction problem. Thirdly, we implement a scheme to select the mostinformative of those features, contributing to a better generalization performance, andoffering useful insights. Two algorithms are used to that end: one based on random forestsand one employing a sparse logistic regression approach. Finally, once the features areselected, we implement, analyse, and compare several ML architectures (from classicalclassifiers to deep learning) to predict the delay. While the focus of the comparison isprediction accuracy, metrics such as sample and computational complexity are alsodiscussed. Numerical experiments are drawn from the US domestic market for the year 2018,when more than 7 million flights between 358 airports were flown. The designedprocessing/learning pipeline reveals interesting insights and achieves better predictionresults than the state of the art. The results confirm that air traffic delay prediction is achallenging problem, mainly because the delay is extremely airport-dependent and the datais highly unbalanced (i.e., only a small percentage of flights are noticeable delayed), andidentify worth-pursuing future lines of work.