Prediction and Uncertainty Quantification of Daily Airport Flight Delays

Thomas Vandal, Max Livingston, Camen Piho, Sam Zimmerman
Proceedings of The 4th International Conference on Predictive Applications and APIs, PMLR 82:45-51, 2018.

Abstract

One in four commercial airline flights is delayed, inconveniencing travelers and causing large financial losses for carriers. The ability to accurately predict delays would make travelers’ lives easier and save airlines money. In this work, we approach the problem of predicting flight delays using a Variational Long Short-Term Memory (LSTM) model. The model is trained to predict aggregate daily delays for U.S. airports using a combination of continuous and discrete variables, including weather, airport characteristics, and congestion. Monte Carlo Dropout, a Bayesian Deep Learning technique based on variational inference, is incorporated to provide planners with a well-calibrated prediction interval. We show that our Variational LSTM results in an average median absolute error of 5.8 minutes per day across 123 airports in the United States. Moreover, results show that predictive uncertainty is well explained through a calibration analysis.

Cite this Paper


BibTeX
@InProceedings{pmlr-v82-vandal18a, title = {Prediction and Uncertainty Quantification of Daily Airport Flight Delays}, author = {Vandal, Thomas and Livingston, Max and Piho, Camen and Zimmerman, Sam}, booktitle = {Proceedings of The 4th International Conference on Predictive Applications and APIs}, pages = {45--51}, year = {2018}, editor = {Hardgrove, Claire and Dorard, Louis and Thompson, Keiran}, volume = {82}, series = {Proceedings of Machine Learning Research}, month = {24--25 Oct}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v82/vandal18a/vandal18a.pdf}, url = {https://proceedings.mlr.press/v82/vandal18a.html}, abstract = {One in four commercial airline flights is delayed, inconveniencing travelers and causing large financial losses for carriers. The ability to accurately predict delays would make travelers’ lives easier and save airlines money. In this work, we approach the problem of predicting flight delays using a Variational Long Short-Term Memory (LSTM) model. The model is trained to predict aggregate daily delays for U.S. airports using a combination of continuous and discrete variables, including weather, airport characteristics, and congestion. Monte Carlo Dropout, a Bayesian Deep Learning technique based on variational inference, is incorporated to provide planners with a well-calibrated prediction interval. We show that our Variational LSTM results in an average median absolute error of 5.8 minutes per day across 123 airports in the United States. Moreover, results show that predictive uncertainty is well explained through a calibration analysis.} }
Endnote
%0 Conference Paper %T Prediction and Uncertainty Quantification of Daily Airport Flight Delays %A Thomas Vandal %A Max Livingston %A Camen Piho %A Sam Zimmerman %B Proceedings of The 4th International Conference on Predictive Applications and APIs %C Proceedings of Machine Learning Research %D 2018 %E Claire Hardgrove %E Louis Dorard %E Keiran Thompson %F pmlr-v82-vandal18a %I PMLR %P 45--51 %U https://proceedings.mlr.press/v82/vandal18a.html %V 82 %X One in four commercial airline flights is delayed, inconveniencing travelers and causing large financial losses for carriers. The ability to accurately predict delays would make travelers’ lives easier and save airlines money. In this work, we approach the problem of predicting flight delays using a Variational Long Short-Term Memory (LSTM) model. The model is trained to predict aggregate daily delays for U.S. airports using a combination of continuous and discrete variables, including weather, airport characteristics, and congestion. Monte Carlo Dropout, a Bayesian Deep Learning technique based on variational inference, is incorporated to provide planners with a well-calibrated prediction interval. We show that our Variational LSTM results in an average median absolute error of 5.8 minutes per day across 123 airports in the United States. Moreover, results show that predictive uncertainty is well explained through a calibration analysis.
APA
Vandal, T., Livingston, M., Piho, C. & Zimmerman, S.. (2018). Prediction and Uncertainty Quantification of Daily Airport Flight Delays. Proceedings of The 4th International Conference on Predictive Applications and APIs, in Proceedings of Machine Learning Research 82:45-51 Available from https://proceedings.mlr.press/v82/vandal18a.html.

Related Material