PROJECT 1

DISTRIBUTION AND ASSIGNMENT

William H. Hsu, Shing Chang, 
 Kansas State University

Proposed Grant $114,957

Abstract:
The contributions of the proposed research on simulation-based monitoring are: (1) advances in two types of probabilistic predictive models (dynamic Bayesian networks and temporal artificial neural networks) and algorithms for learning them from personnel history; (2) a prognostic monitoring algorithm that uses these models to predict the migration of skills across organizations given personnel distribution;

 and (3) an algorithm for skill set optimization by dynamic programming given data (records on personnel management and recent migration histories), predictions (available manpower and requirements), and quantitative constraints (skill set needs from the organizational requirements specification). Previous algorithms for intelligent pre-filtering and assignment of personnel are based on assumptions that are insufficient for reasoning under uncertainty using observed historical data. Our proposed simulation-based system will use the above state-of-the-art data mining technologies to build probabilistic models from this data. The experimental focus of this research is to compare the quality of these probabilistic predictive models for optimization and decision support with that of those produced by traditional multi-objective decision making models.

Three novel and innovative aspects of this research are as follows. First, the proposed system will be able to elicit and use objectives provided by the field recruiter and screening manager during skill set specification. The research challenge is to develop a simulation-based model that can compile these total force skill objectives into constraint-based knowledge to guide the accumulation of required skill sets, through targeted recruiting and pre-filtering of candidate Sailors. Second, the research problem of developing an intelligent system for skill set allocation leads to our proposed method of simulation-based monitoring to produce and display continuous predictions of personnel migration and skill set demand. The research challenge is to incorporate records on demographics and qualifications of Sailors into a temporal uncertain reasoning model, such as a Bayesian network or neural network, that can monitor the migration of skill sets through an organization over time. Third, this research will evaluate the utility of this model in generating predictions as input to skill set optimization algorithms that produce recommendations on the placement (distribution and assignment) of the qualified Sailor, given specified needs. We propose to compare adaptive programming techniques – simple dynamic programming and multi-attribute, multi-objective decision making (MADM and MODM) algorithms – for this back end.