"Predictive Analytics for Military Construction Overruns: A Machine Lea" by William D. Hunter

Date of Award

3-2024

Document Type

Thesis

Degree Name

Master of Science in Engineering Management

Department

Department of Systems Engineering and Management

First Advisor

Brent T. Langhals, PhD

Abstract

This research focused on analyzing cost and schedule overruns in a specific set of Air Force MILCON projects. Using data from the Built Infrastructure Common Operating Picture (BICOP), several advanced machine learning techniques, including LASSO regression and Random Forest, were applied to uncover patterns and build predictive models. The work sought to enhance the decision-making capabilities and efficiency of project managers within the Air Force through data-driven insights. A key finding from the models indicated that location was the most important factor in influencing a project’s tendency toward cost and schedule overrun. The best model for predicting schedule overrun achieved an accuracy of 82%, which was 20% higher than the no-information rate. In contrast, the best model in predicting cost overrun attained an accuracy of 70%, which was 11% higher than the no-information rate. Although the predictive models demonstrated some level of predictive capability and exceeded the no-information rate, they were not consistently accurate enough to be recommended as an aid for decision-making. The findings suggest that while machine learning can provide valuable insights into factors influencing project overruns, further refinement and testing are needed before these models can be considered reliable for use to flag early signs of potential construction project delays or cost increases.

AFIT Designator

AFIT-ENV-MS-24-M-132

Comments

A 12-month embargo was observed for posting this work on AFIT Scholar.

Distribution Statement A, Approved for Public Release. PA case number on file.

Share

COinS