Theses and Dissertations

ORCID

https://orcid.org/0009-0009-5463-7772

Advisor

Ma, Junfeng

Committee Member

Bian, Linkan

Committee Member

Tajik, Nazanin

Committee Member

Drzymalski, Julie

Date of Degree

5-16-2025

Original embargo terms

Immediate Worldwide Access

Document Type

Dissertation - Open Access

Major

Industrial and Systems Engineering

Degree Name

Doctor of Philosophy (Ph.D.)

College

James Worth Bagley College of Engineering

Department

Department of Industrial and Systems Engineering

Abstract

Predicting air passenger volumes is crucial for airports and airlines seeking to reduce costs and enhance profitability. Accurate forecasting enables better planning and efficiency improvements within the air transport industry. This study applies LSTM, ARIMA and HW to U.S. air passenger datasets. Each analysis shows a methodology for predicting air passenger volumes across airports, airlines and across airports and airlines simultaneously. ARIMA was found to have limited applicability, since only a subset of the datasets was stationary. LSTM and HW were applicable to all airlines and ARIMA was applicable to no airlines. LSTM had less error compared to HW at the airline level, indicating LSTM is a more accurate predictor of air passenger volumes at the airline level. LSTM had significantly lower MAE values than both HW and ARIMA at the airport level which reinforces that LSTM is superior at forecasting air passenger volumes at the airport level. LSTM generally exhibited lower mean absolute error (MAE) values for the datasets spanning airlines and airports. In addition, LSTM had the lowest error rate across each of the four tiers of data. This means that utilizing LSTM to predict U.S. air passenger enplanements improves accuracy of forecasting when compared to HW and ARIMA. The results suggest that LSTM’s ability to merge long and short-term data dependencies is most influential when predicting air passenger enplanements.

Sponsorship (Optional)

W.L. Gore & Associates

Download

Share

COinS