A model for predicting the repair costs of U.S. Navy inventory items

Abstract

This research was initiated due to a report claiming that The U.S. Navy significantly overestimated repair prices in a Performance Based Logistics (PBL) reward to a contractor. The purpose of this thesis is to develop a model for improving the prediction of repair price for U.S. Navy inventory items. The thesis examines several prediction methodologies, including a ratio-estimator prediction method that is a modification of the methodology currently in use, as well as regression analysis. In contrast to the ratio-estimator approach, regression is able to utilize a wide range of predictor variables, several of which are evaluated in the thesis research. Results of this analysis reveal that a regression model with logarithmic transformations yields more accurate predictions of repair prices than the current methodology. This improvement is seen especially for items that have the highest replacement price. One feature of the proposed regression-based methodology is that predicted repair prices for the most expensive items are substantially lower than with the current methodology. In the case where that prompted the thesis research, the overstatement of benefit from the PBL would have been reduced by about 30 million under the proposed methodology.