Automated forecasting of replacement part requisitions

This thesis develops a forecasting model to predict replacement part requisitions on a monthly or quarterly basis for the Naval Surface Warfare Center (NSWC) in Crane, Indiana. The replacement parts are components in the AN/SLQ-32 Electronic Warfare (EW) system, which is apart of the Navy's Ship Self Defense System. The system provides electronic support for radio frequency bands in the detection of incoming anti-ship missile (ASM) terminal guidance radars.; A number of forecasting methods were investigated for the monthly and quarterly historical part requisition data. A multiple linear regression analysis resulted in the best forecasting model which includes both past requisition demand and steaming hours (as a measure of fleet operational hours) to predict the quarterly requisition demand. A regression model with an average of 5--6 quarter lags in the steaming hours was the best forecasting model. The lags correspond with the estimated work up cycle plus deployment periods of ships. Therefore, steaming hours seems to be a good predictor of quarterly part requisitions both mathematically within the regression model and practically.; The current forecasting method is a single exponential smoothing model with an unknown smoothing constant (alpha) model parameter. For every studied part, there is at least a 50% or greater improvement in the forecast accuracy with the multiple regression models than the current single exponential smoothing forecasting models. Even after applying standard parameters (i.e. alpha = .20 and a 4 quarter lag) for both models to all parts, there is a significant, at least a 26% or greater, improvement in the forecast accuracy with the multiple regression models than the current single exponential smoothing models.