![IconGems-Print](images/IconGems-Print.png)
619 – Topics in Defense and National Security
A New Military Retention Prediction Model: Machine Learning for High-Fidelity Prediction
James Bishop
Institute for Defense Analyses
Alan Gelder
Institute for Defense Analyses
Michael Guggisberg
Institute for Defense Analyses
Joe King
Institute for Defense Analyses
Julie Lockwood Pechacek
Institute for Defense Analyses
Cullen Roberts
Institute for Defense Analyses
The Department of Defense (DoD) is the largest employer in the United States of America with 2.15 million service members in 2019. The DoD must anticipate retention of service members to effectively perform its mission. A team from the Institute of Defense Analyses began development of the Retention Prediction Model (RPM), an application of a feed-forward neural network on a novel dataset. The RPM is a tool that predicts retention of service members in the DoD and can be used to inform DoD leadership about anticipated retention at the service member level. Prediction on an out-of-sample set results in a concordance no worse than 0.78 for any given year or 0.73 for the restricted mean survival time.