Dissertation/Thesis Abstract

Predicting Annual Flying Hours Under Conditions of Uncertainty
by Schneider, Rose, M.S., Southern Illinois University at Edwardsville, 2017, 96; 10269250
Abstract (Summary)

The United States Transportation Command (USTRANSCOM) is responsible for managing the global defense transportation network and creating movement plans in response to various world events, such as military contingencies, humanitarian events, natural disaster response, etc. Historically, USTRANSCOM has struggled with estimating the budget for future fiscal years. In the recent past, the Command has overestimated annual flying hours needed by 25% and underestimated by as much as 402%. To remedy this situation, USTRANSCOM has tasked The MITRE Corporation with performing analysis on past historical flight data and creating an application that generates forecasts for the next fiscal year, which will then drive the transportation budget.

Historical flight data has been gathered from 2010-2015. As well, pertinent humanitarian effort data, notably meteorological and seismological events, has been captured. Much work has been done to cleanse and aggregate the data to prepare it for analysis. Using the programming language R and open source platform R Studio, a program was created that ingests each data set and produces forecasts using exponential smoothing and autoregressive integration moving average methods. The program is flexible and displays the “best” forecast based on user-selected criteria. Due to restrictions on government data, a representative data set (annual flights in and out of Los Angeles International Airport from 2006-2016) serves as a case study.

Using the R library Shiny, the forecasting program was converted to an easily-deployable web application, which will serve as a proof-of-concept design for eventual delivery to USTRANSCOM. This application is intended to work in tandem with a data visualization program that is being developed by The MITRE Corporation. Users will be able to select and customize which historical data set they would like to forecast. This data is then sent to the application and forecasts are generated automatically. Thus, the forecasting tool is intended to be completely dynamic—regardless of the data being sent, the application clearly graphs the historical data, generates and displays forecasts using various forecasting methods, and picks the “best” forecast based on user choice (with mean absolute squared error serving as the current default). This application is intended to easily provide visualizations on past historical data and assist USTRANSCOM users in making informed decisions on future transportation needs.

Indexing (document details)
Advisor: Chen, Xin
Commitee: Eneyo, Emmanuel S., Onal, Sinan
School: Southern Illinois University at Edwardsville
Department: Mechanical and Industrial Engineering
School Location: United States -- Illinois
Source: MAI 56/04M(E), Masters Abstracts International
Subjects: Industrial engineering
Publication Number: 10269250
ISBN: 978-1-369-79692-6
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy