Forecasting Air Force Logistics Command Second Destination Transportation: An Application of Multiple Regression Analysis and Neural Networks
Abstract
The Air Force Logistics Command currently uses a simple linear regression model to forecast overseas Second Destination Transportation (SDT) general cargo tonnage requirements for specific geographical areas. The independent variable for the model is the total flying hours for each geographical area while the dependent variable is the general cargo SDT tonnage requirement. This thesis explored the use of a multivariable approach for developing multiple regression and neural network models which was based on the breakout of the total flying hour variable into separate aircraft flying hours and the addition of military population variables. The purpose of this research was to develop multiple regression and neural network models for predicting Pacific and European Military Airlift Command and Military Sealift Command general cargo tonnage requirements that were more accurate forecasting models than the simple regression forecasting models presently used by AFLC/DSXR. Once the models were developed, the multiple regression and neural network models were compared to determine which type of model was statistically more accurate. Neural networks are an adaptive information processing system loosely based on the information processing capability of the human brain that mathematically develops associations between particular independent and dependent variables. Recent research indicates neural network models are an alternative to conventional mathematical techniques for solving problems that do not have a well defined model or theory. (KR)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1990
- Accession Number
- ADA229629
Entities
People
- Kevin R. Moore
Organizations
- Air Force Institute of Technology