Student tutors for the course Data Modelling wanted (winter term 2021/22)

IT-Project Data Souvereignty in winter termin 2021/22

Business Intelligence: Washing Gold in Times of Information Overload

See all news.



You can find us here.

Predicting Flight Delay Risk Using a Random Forest Classifier Based on Air Traffic Scenarios and Environmental Conditions

Authors: M. Bardach, E. Gringinger, M. Schrefl, C. Schütz
Paper: Schu20b (2020)
Citation: Proceedings of the 2020 AIAA/IEEE 39th Digital Avionics Systems Conference (DASC), October 11-15, 2020, San Antonio, Texas, U.S.A., IEEE Computer Society, DOI: 10.1109/DASC50938.2020.9256474, 2020.
Resources: Copy  (In order to obtain the copy please send an email with subject  Schu20b  to

Abstract (English): A reduction of delay costs can be achieved through more adaptable flight planning, which hinges on accurate prediction of delays. In order to counteract the expected delay of flights, air traffic control may adapt flight plans through slot swapping, opening another runway, or changing the runway configuration, for example. Environmental conditions and external events such as runway and airspace closures may render a flight plan obsolete, which must be taken into account when aiming to reduce delay. Air traffic control must recognize changes in the environment and external events such as runway and airspace closures as early as possible in order to adapt flight plans accordingly and avoid delays. Current systems employed by air traffic control do not sufficiently leverage the multitude of available data for the detection of upcoming congestion and, consequently, flight delays. Therefore, flight plans are not adapted fast enough in air traffic scenarios with potentially high delay. In this paper, we aim to predict the risk class of an air traffic scenario based on the expected cost of the delays, and considering information about environmental conditions and external events. In particular, we present a random forest classifier for Atlanta International Airport, which achieves an accuracy of 82.5% for the highest and thus most important risk classes. The development of similar classifiers for other airports may help air traffic control to more accurately predict scenarios with high congestion, and counteract accordingly in the future.