Predicting Flight Demand under Uncertainty

Abstract

The motivation of this study originates from the need for the United States Transportation Command (USTRANSCOM) to estimate the annual flying hours budget for future fiscal years. USTRANSCOM is responsible for managing the global defense transportation network and creating movement plans in response to various world events. Recently, USTRANSCOM has overestimated annual flying hours by 25% and underestimated by 402%. This article presents analyses of historical flight data and a methodology to forecast annual flying hours. A main contribution of this article is the meticulous coupling of the forecasting methodology and intricacies of a complex and challenging real-world problem. The methodology may be extended to applications where historical data are used to predict future demand and requirements under uncertainty. An application using the programming language R and open source platform R Studio is developed to ingest data and produce forecasts using the methodology developed in this article. Due to data classification concerns, a representative data set of annual flights in and out of Los Angeles International Airport from 2006–2016 is used as a case study to validate the methodology for public release.