Improved Enlisted Retention, Attrition and Loss Prediction Methodologies - Cyber Academic Group
This research will explore the use of multiple statistical analysis approaches to project enlisted Navy sailor retention behavior. Approaches to investigate include regression analysis, “credit” scoring, big data techniques, machine learning and random forests. The methods can then be compared for feasibility and effectiveness and potential for future use.The goal of this discovery effort is to explore the relationship between descriptive factors of enlisted sailors and their propensity to attrite, leave or reenlist at or before their first term of enlistment. OPNAV N1 (CNP – Chief of Naval Personnel) has long desired a more accurate means of predicting future loss and retention values, which would assist strength planners, policy makers, community managers, production line managers and training capacity planners to produce more accurate, defendable and actionable forecasts. This proposal is for an internal N81 study, in collaboration with the Naval Postgraduate School, to investigate and identify those factors (race, ethnicity, gender, education level, dependency status, geographic location, geographic economic factors, RCN, etc.) that impact retention to develop loss probabilities/survival rates for use in predicting the future behavior of a given enlistment cohort. By quantifying the relationship between known historical data and sailor behavior, Navy can gain greater accuracy in strength planning, budget generation, and the manpower programming rates used throughout the PPBE process.
Office of the Chief of Naval Operations