Date of Award

3-23-2018

Document Type

Thesis

Degree Name

Master of Science in Operations Research

Department

Department of Operational Sciences

First Advisor

Brian J. Lunday, PhD.

Abstract

The Joint Distribution Processing Analysis Center (JDPAC) of the United States Transportation Command (USTRANSCOM) regularly forecasts the demand of USTRANSCOM assets required by geographic and combatant commanders. These demands are subject to fluctuations due to unforeseen circumstances such as war, conflict, natural disasters, and other calamities requiring the presence of military personnel. This study evaluates the use of exponential state space smoothing, ARIMA, and Regression with ARIMA errors models to forecast the number of military personnel expected in each country, for a test set of countries of interest to USTRANSCOM and which manifest a high degree of variability in the anticipated number of troops each year. The expectation by USTRANSCOM is that accurate forecasts for the number of military personnel in each country can be leveraged to develop alternative transportation workload forecasts of demand of USTRANSCOM assets. There was not a single model that performed best for all countries and branches of service. Each model was analyzed via the traditional 80/20 forecasting evaluation metric as well as a two-year horizon cross-validation metric. The exponential smoothing model with a high level of α performed quite well for many of the models, indicating that perhaps simpler models will still provide accurate forecasts. Further research is needed to determine whether incorporating forecasts of military personnel will improve the ability to forecast demand of USTRANSCOM assets.

AFIT Designator

AFIT-ENS-MS-18-M-162

DTIC Accession Number

AD1056423

Download

Share

COinS