Multivariate Manpower, Personnel and Training (MPT) Modeling and Management System

The Navy would greatly benefit from a computerized system that identifies environmental variables/personnel characteristics that predict accession, trainability, assignment performance, and retention of potential recruits/current personnel. A Bayesian network (BN) methodology is ideally suited to identifying/modeling predictive variables and their impacts over time. In Phase I, we propose using BNs in a multivariate statistical analysis of variables/characteristics that are important to predicting behavior for ten Navy/Marine ratings. The resultant computerized BN model will capture relationships between variables/characteristics and accession, training, performance, and retention behavior of current or future Navy personnel, so that impacts of proposed/hypothesized changes in environmental variables (such as Navy personnel policies), or personnel characteristics could be quickly evaluated. We will also anced human factors engineering principles. Since determining optimal policies by trial and error with the BN model could be cumbersome, our Phase I Option will develop a Genetic Algorithm software component to interface with the BN model to automatically explore and timal policies/characteristics based on user specified optimality criteria (e.g., retention). The option will also develop a plan for validating the system, scaling up to many more ratings, and identifying tradeoff opportunities in Phase II.