PREDICTING U.S. ARMY ENLISTED ATTRITION AFTER INITIAL ENTRY TRAINING USING RANDOM SURVIVAL FORESTS
Loading...
Authors
Lazzarevich, Nicholas R.
Subjects
Army
enlisted
attrition
survival analysis
machine learning
medical data
predict
random forests
random survival forests
Person-Event Data Environment
enlisted
attrition
survival analysis
machine learning
medical data
predict
random forests
random survival forests
Person-Event Data Environment
Advisors
Buttrey, Samuel E.
Date of Issue
2022-03
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
The U.S. Army requires models that predict the proportion of post–Initial Entry Training (IET) soldiers who complete their initial term of service, and which assess the risk of attrition prior to completion at various points during these terms. The Army struggles to access sufficient recruits to maintain approved personnel levels due to economic competition and a shrinking population of candidates who are both willing and eligible for recruitment. Roughly 24% of soldiers who complete IET fail to complete their initial term of service. Modeling post-IET attrition and identifying factors that contribute to attrition will allow the Army to access soldiers with lower risk of attrition and assess policies to address attrition throughout the initial term. Continuing work done by Devig in 2019 with survival analysis, this research utilizes the randomForestSRC R package by Ishwaran and Kogalur in 2020 to build a series of random survival forests, allowing us to approximate effects of time-varying covariates (TVC). This research uses data stored in the Person-Event Data Environment and consists of demographics, deployments, medical readiness, and initial entry data. Using fiscal year (FY) 2010 as a training set and FY 2011 as a test set, we find that two of the top 10 predictors are medical while the rest are demographic, and four are TVC. The final models perform well for predicting cohort attrition at various points during the first term, but not for attrition of individuals.
Type
Thesis
Description
Series/Report No
Department
Operations Research (OR)
Organization
Identifiers
NPS Report Number
Sponsors
Funder
Format
Citation
Distribution Statement
Approved for public release. Distribution is unlimited.
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.