Although most airmen successfully complete initial skills training, about 10 percent are eliminated and either separated from the United States Air Force or reclassified into other specialties. Given recent increases in reclassification, the authors set out to identify factors that affect initial skills training success and determine how classification and reclassification processes can be improved.
U.S. Air Force Enlisted Classification and Reclassification
Potential Improvements Using Machine Learning and Optimization Models
Download eBook for Free
|PDF file||7.1 MB||
Use Adobe Acrobat Reader version 10 or higher for the best experience.
Purchase Print Copy
|Add to Cart||Paperback142 pages||$40.00||$32.00 20% Web Discount|
- What policies does the USAF currently follow for classification and reclassification of airmen, and how can both outcomes be improved?
- What factors contribute to IST success and failure?
- What ML approaches are best suited to identifying optimal AFS assignments, thus reducing the need for reclassification?
- How can reclassification be improved in order to achieve better training and career outcomes and reduce costs?
Recent trends in initial skills training (IST) for Air Force specialties (AFSs) indicate that the number of United States Air Force (USAF) enlisted personnel reclassified into other occupational specialties has increased in recent years, with a steady rise having occurred between fiscal years 2013 and 2017. Career field reclassification can result in a wide range of negative outcomes, including increased costs, delayed manning, training schedule challenges, and decreased morale. To understand and address the challenge of IST reclassification, the authors considered options for improving processes to classify and reclassify enlisted active-duty, non–prior service airmen for IST. In this report, they outline key findings from a 2019 study that employed qualitative and quantitative analyses, including machine learning (ML) models, to assess predictors of IST success (and failure). They also describe their test of an optimization model designed to identify opportunities for revising reclassification decisions in order to not only reduce the numbers of reclassified airmen but also to achieve greater job satisfaction and productivity for airmen and improve USAF retention rates.
IST classification is designed to optimize training success but not other important outcomes
- With average graduation rates of 95 percent, further efforts to optimize training success might yield minimal gains.
- However, ML models may be effective in predicting early separation and reenlistment.
Increasing the number of relevant variables can increase the accuracy of ML predictions
- The types of ML models used differed by less than 1 percent in predicting outcomes.
- Expanding the set of predictor variables in the ML models generally decreased prediction errors by approximately 5 percent.
Reclassification is a manual process and can be optimized to achieve different outcomes
- USAF might not be using the minimal cost solution for IST reclassifications or the solution that produces the maximum number of positive training and career outcomes.
- Reclassifying airmen with optimization models to achieve optimal training and career outcomes will increase the costs of reclassification.
- Alternative solutions that achieve slightly better training and career outcomes while also reducing the costs currently associated with reclassification are also possible.
Focus group discussions with airmen in IST for selected AFSs identified factors contributing to IST success and challenges and identified suggested improvements
- Airmen characteristics (e.g., motivation) and prior experiences (e.g., education), supportive instructors, and study groups contribute to IST success.
- IST challenges involve both airmen characteristics and the training base environment.
- Improvements cover such areas as prior knowledge of AFSs and what to expect from IST, curriculum design, non-IST requirements, and dormitory arrangements.
- Expand the set of predictors used in USAF enlisted classification by retaining databases concerning qualifications for and outcomes of IST, requiring recruits to complete vocational assessments and recruiters to provide information about IST and AFSs, systematically collecting information about job requirements, developing a biodata instrument to be completed by all enlisted recruits, and using peers and instructors to rate airmen's personalities.
- Expand the set of outcomes used in USAF enlisted classification by defining and systematically measuring outcomes beyond those associated with IST success and by monitoring the moving average for graduation by specialty.
- Improve data quality, comprehensiveness, and access so that ML models can provide accurate and useful predictions.
- Update the classification and reclassification processes to optimize not only IST success but also job match, thus improving performance and career satisfaction.
- Address challenges in such areas as ethics and privacy, interpretability of ML models, and model performance before implementing any ML model.
Table of Contents
Introduction and Background
Air Force Classification and Reclassification Processes
Data Available for Predicting Air Force Training and Career Outcomes
Models to Predict Success
Optimization Model for Reclassifying Training Eliminations
Airmen Experiences in Initial Skills Training for Select Specialties
Conclusions and Recommendations
Defining and Measuring Success in Personnel Selection
Descriptive Statistics and Analytic Modeling Results
Optimization Model Methodology
Focus Group Methodology
Research conducted by
The research described in this report was sponsored by the Director of Operations and Communications, Headquarters Air Education and Training Command, USAF and conducted within the Workforce, Development, and Health Program of RAND Project AIR FORCE (PAF).
This report is part of the RAND Corporation Research report series. RAND reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for research quality and objectivity.
This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited; linking directly to this product page is encouraged. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial purposes. For information on reprint and reuse permissions, please visit www.rand.org/pubs/permissions.
The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.