Knowledge representation for model libraries

Abstract

A knowledge-representation scheme is proposed for a library of complex models with an emphasis on knowledge of management-science paradigms. The scheme supports the model type, the model template, and the model instance levels. At the model-type level, reasoning about assumptions, mathematical form, inputs, outputs, and solvers is supported. The library maintains a taxonomy of model types based on these features. At the model template and model instance levels, reasoning about model details and business knowledge incorporated into the model are supported. An overview of the library operators is provided, and the assumption matching operator, which permits inexact searching, is described in detail.<>