Navy recruit attrition prediction modeling

Abstract

This study develops a model to predict a potential recruit’s likelihood or probability of surviving through the first term of his or her enlistment based on information available to recruiters. This model is compared and contrasted with the predictive ability of the current Navy Recruit Quality Matrix, which classifies recruits into three categories: A, B, and Cu. The data used for this study was from information obtained from recruits at time of accession during fiscal year 2006–2013. We found evidence that there are other recruit characteristics identified at time of recruitment, other than his or her Quality Matrix categories, which may indicate recruits who are at greater risk of attriting. Some of these variables, such as Armed Forces Qualification Test percentile score, education, and body mass index, might contribute to developing a recruit-screening tool. Others, such as gender, will not be appropriate for such use. However, the estimated probabilities computed from the logistic regression model of this thesis can be used to identify subsets of recruits who have a high probability of completing the first term that would normally not be identified through the Navy Recruit Quality Matrix.