"An interactive optimization component for solving parameter estimation and policy decision problems in complex simulations"

The increased use of large, nonlinear and dynamic models to represent real world phenomena has resulted in complex parameter estimation and policy decision problems. This paper describes an interactive optimization component, an integrated set of minimization techniques, to assist in the efficient solution of these problems. The component was developed with particular consideration given to computer execution requirements, interaction flexibility, problem solution accuracy, and component/computer compatibility. A pattern recognition sub-component used for examining model behavior and an optimization sub-component used for multi-level numerical searches combine to provide good problem solutions for broad classes of problems. Test results of component application for a Nigerian beef industry model and a single-commodity market model are discussed and analyzed. Finally general conclusions and further extensions of the research are presented.