Modeling and Forecasting the Demand for Aircraft Recoverable Spare Parts

This report explores issues in forecasting and modeling the demand for aircraft recoverable spare parts to improve the Air Force's estimation of spares and repair requirements over quarterly, annual, and longer planning horizons. Specifically, it demonstrates the utility of approaches that account explicitly for nonstationarity and their superiority over current methods used by the Air Force Materiel Command for these purposes. The authors recommend using a weighted regression, a special case of the Kalman filter, for forecasting demand for high-demand items. This approach is a logical extension of Bayesian statistics, which explicitly accounts for nonstationarity in stochastic processes, assigning greater weight to more recent than to less recent demands. Coupled with an improved approach to variance estimation that assigns greater uncertainty to longer planning horizons than to shorter ones, this holds the promise of reducing the cost of spares investments while achieving adequate levels of system performance.