A New Military Retention Prediction Model: Machine Learning for High-Fidelity Forecasting

June, 2019
IDA document: D-10712
FFRDC: Systems and Analyses Center
Type: Documents
Division: Strategy, Forces and Resources Division
Julie Pechacekn, Alan Gelder, Cullen Roberts, Joe King, James Bishop, Michael Guggisberg, Yev Kirpichevsky See more authors
Using machine learning algorithms and 18 years of data, we predict individual-level attrition among active duty personnel in all military Services, with hold-out sample prediction accuracies typically exceeding 70%. Importantly, our methodology accommodates both right and left-censoring of observed career paths, and significantly outperforms traditional survival analysis. Using these individual-level predictions, we generate aggregate predicted force profiles which closely align with historical actuals. This and other features offer a rich slate of observations for further empirical analysis, and suggest new policy levers for managing attrition