Predicting U.S. Army Enlisted Attrition After Initial Entry Training Using Survival Analysis
Abstract
For the first time since 2005, the U.S. Army fell short of its recruiting goal in 2018 by about 6,500 recruits. A strong economy and an increasing pool of recruit candidates who require a waiver to enlist add to the Armys recruitment troubles. Mental health issues, obesity, and other medical issues have become barriers that disqualify recruits from enlisting. For those who are eligible, they complete a training period called Initial Entry Training (IET). After finishing IET, many soldiers do not finish their first-term service obligation. This research continues the research conducted by Speten (2018) on post-IET attrition, with the added benefit of having medical data available in the Person-event Data Environment (PDE), a secure, virtual environment with a database that provides information on manpower, service, personnel, and medical data. Currently, no research has been conducted that uses detailed medical information to predict post-IET attrition. To estimate the expected number of soldiers who attrite at a specific time post-IET and prior to the end of their first-term obligation, we construct survival tree models using time-varying and time-constant covariates. We find several medical covariates that are important in forecasting attrition including dental readiness and hearing readiness. The effectiveness of the models is assessed on independent test sets. They perform well in predicting expected number of attrition, but not in predicting individual soldier attrition.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2019
- Accession Number
- AD1080192
Entities
People
- Aaron L. Devig
Organizations
- Naval Postgraduate School