Automating Model Formulation for Decision Support

Abstract

Many still believe that model formulation is a by-product of selecting a proper problem or tool representation technique. The recent trend is to divorce model formulation from model execution and study it separately. By automating the model formulation process, we create an internal representation of the formulated model. In this paper, we discuss an architecture that has been developed to support the model formulation process. Model formulation in our context is defined as the process that helps transform a problem description into a solvable algebraic notation. The problems being focused on are transportation problems in the production planning domain. As a first step, a brief overview to a cognitive model developed by the authors for the model formulation process is discussed. Design features extracted from this cognitive model for facilitating the automation of the model formulation process are presented. The architecture is based on the blackboard paradigm and attempts to capture the nuances of the cognitive model. A prototype has been developed using this architecture, which bears out the viability of the approach.