Analysis of the Air Force Logistics Command (AFLC) Forecasting Method for Predicting Second Destination Transportation (SDT).

摘要

This research was conducted to analyze the Air Force Logistics Command (AFLC) forecasting method for predicting Second Destination Transportation (SDT) tonnage with the regional flying hour program. This thesis had two main objectives: (1) validate the current forecasting method used for computing tonnage estimates to derive SDT budget requests, and (2) if the current method's validity was not supported, develop a new forecasting model, using the same input data, that would produce more accurate and reliable tonnage estimates. Analyzing graphs of the four different data sets researched in this thesis and then conducting a statistical test on the flying hour parameter for each set, it was determined that the current method employed to forecast SDT tonnage was statistically invalid for two of the four sets. This determination was made due to the fact that the flying hour parameter changed during the iterative regression process used. This change implied that SDT tonnage and flying hours were not linearly related. Box-Jenkins (BJ) time series forecasting models for each data set were built and provided accurate and valid forecasts. For the MAC SDT time series, the BJ models were more accurate than the current method. The BJ models for the MSC SDT time series, although marginally less accurate than the current method, were valid, whereas the current method for these two series was statistically invalid. Theses.