AUTOMATED SUPPORT FOR FORMULATING LINEAR PROGRAMS

Abstract

Most research in mathematical programming has been concerned with efficient computational algorithms. However, there is increasing interest in developing automated techniques for supporting the modeling process. This paper describes a new kind of interface for formulating linear programming models and explains the inference process used to translate problem specifications into algebraic formulations. The main idea underlying the design of the interface is to change the specification language to a graphical rather than a mathematical notation. The inference process involves the generation of algebraic terms and their subsequent combination into constraint equations. This relies on the syntactic relationships among indices and a knowledge of the physical entities that they represent. An advantage of the approach is that it facilitates the reuse of model components from previous models. The ideas discussed in this paper have been incorporated in a prototype system.