PERFORMANCE OF RANDOM SURVIVAL FORESTS WITH TIME-VARYING COVARIATES IN PREDICTION OF U.S. ARMY ENLISTED ATTRITION COMPARED TO TRADITIONAL MANPOWER ANALYSIS METHODS
Loading...
Authors
Rooney, Connor A.
Subjects
survival analysis
attrition
random survival forest
RSF
time-varying covariates
T-VC
forecasting attrition
army
random survival forest with time-varying covariates
TV-RSF
person-event data environment
PDE
left-truncated right-censured
LTRC random forest
statistical learning
machine learning
DOD manpower
forecast manpower
survival trees
logistic regression
random forests
attrition
random survival forest
RSF
time-varying covariates
T-VC
forecasting attrition
army
random survival forest with time-varying covariates
TV-RSF
person-event data environment
PDE
left-truncated right-censured
LTRC random forest
statistical learning
machine learning
DOD manpower
forecast manpower
survival trees
logistic regression
random forests
Advisors
Buttrey, Samuel E.
Date of Issue
2022-12
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
The importance of identifying qualified candidates and properly forecasting future manpower strength will always be critical to military recruiting and organization. The ability to assess the cross-section of covariates of a cohort of enlistees and forecast manpower strength would allow for improved planning and allocation decisions. We leverage an innovative method of survival analysis—random survival forests (RSF) with time-varying covariates (T-VC)—to predict Army first-term post-Initial Entry Training attrition rates. Using random survival forests with time-varying covariates (TV-RSF) is an emerging method of survival analysis that has not been used in a military manpower setting. Using a Brier Score we compare TV-RSF with three other methods. We illustrate that using a single tree rather than the computationally intensive TV-RSF may suffice for predicting future year attrition. We also illustrate that TV-RSFs outperform traditional classification methods (logistic regression, random forests) that only account for yearly changes in T-VCs.
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.