A Preliminary Analysis of Alternative Forecasting Techniques for the Standard Base Supply System (SBSS).

摘要

The purpose of this research is to investigate alternative approaches to forecasting demand for expendable items in the Standard Base Supply System (SBSS). The forecasting models studied include single, double and adaptive exponential smoothing. Samples were selected from Dover AFB, Delaware and analysis of the various smoothing models was performed by a FORTRAN program written for each model. Comparison of the forecasting models was made on the basis of forecast error as measured by mean absolute deviation (MAD). The forecast error was also measured for the current forecasting model used by the SBSS. Single expotential smoothing, with small smoothing constants, proved to be the model with the lowest forecast error rate. Program activity was also studied as a possible tool to be used in demand prediction. Flying hours correlated with demand levels for some Federal stock classes. Suggestions for further study are included. (Author)