AgileGIST - a Framework for Iterative Development and Rapid Prototyping of DSS for Combinatorial Problems

Summary Developing model-based decision support systems for combinatorial problems you often encounter the situation that the problem owner is not able to explicate all necessary information to set up the system of constraints and objectives but that he is able and willing to criticise solutions which are obtained from some model implementation. Thus in such scenarios it is necessary for the success of the project to design a development process based on a sequence of models with associated prototype systems and to integrate the problem owner into an iterative and experimental procedure in which (additional) components can be added or dropped from the model base. Heuristics based on indirect search have shown to be suitable and rather powerful in these environments. These problems are typical scenarios in DSS-development and have motivated the development of our new framework: they require new problem-specific representational and algorithmic knowledge which has to be acquired during an initial conceptual phase, exibility rather than absolute accuracy as in literature problems is a dominant system requirement, and, to allow for rapid system development the use of general problem solving capabilities captured in standard coding schemes and metaheuristic strategies is more or less mandatory. In earlier papers we have reported successful applications of the so-called GIST-approach to non-standard real-world decision problems which had been given to us in form of a semi-structured planning task and for which we had to develop a decision support systems (cf.[1],[2]and[3]).In this paper we describe how evolutionary system development using the GISTapproach can be supported by a framework called AgileGIST which has been motivated by two concepts from different areas: mathematical programming languages and the dependency injection framework from software engineering.