PREDICTING METRICS FOR NAVAL SUPPLY SYSTEMS COMMAND WHOLESALE INVENTORY OPTIMIZATION MODEL

Naval Supply Systems Command, Weapons Systems Support relies on discrete-event simulation to generate performance metrics for candidate inventory policies (CIPs), followed by mixed-integer optimization to select the policy mix that best meets enterprise-level objectives. This approach is computationally expensive. This thesis investigates whether key performance metrics, such as expected ready for issue cost (RFIC), fill rate (FR), documents required (DOCS), demand wait time (DWT), and demand wait time for backorder (DWTBO) can be accurately predicted using machine learning models instead of repeated simulation runs. We examine two machine learning approaches: k-nearest neighbors (KNN) and Extreme Gradient Boosting (XGBoost). Both methods train very fast for more than 21,000 items and multiple CIPs by item. These observations are divided into four categories, and the results demonstrate that XGBoost consistently outperforms KNN. Neither method can predict RFIC, DWT and DWTBO with sufficient accuracy to replace the simulation. On the other hand, performance is moderate for DOCS, and very good for FR, with mean absolute errors under 1.0 documents and 0.1 (rate), respectively, across all categories.

Department dc:contributor.department

Operations Research (OR)

Grantor dc:publisher

Monterey, CA; Naval Postgraduate School

Year dc:date.issued

2026

Author dc:creator

• Patterson, Thomas B.

Advisor dc:contributor.advisor

• Salmeron-Medrano, Javier

Statement dc:rights

• This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.

Handle dc:identifier.uri

https://hdl.handle.net/10945/75025

OAI identifier oai:identifier

oai:calhoun.nps.edu:10945/75025