Integrating optimization models and human expertise in decision support tools

摘要

Decision Support literature, specially in the subarea of Model Management, has long recognized that complex problems rarely can be solved by a single model encompassing all problem aspects. Solutions to such problems often require the integration of multiple models each addressing a specific aspect of the problem. Considerable heuristic knowledge from human experts is also required in the solution procedure. In this paper, we present the architecture of a design tool that facilitates interaction among models and a human expert in arriving at a complete solution to a complex problem. A blackboard and an associated truth maintenance system form the basis of computer-based support for such interaction. The architecture of the design tool is independent of the application domain. Its implementation in Turbo Pascal is demonstrated using a problem in distributed computer systems design.