On the use of expert systems in network optimization: With an application to matrix balancing

Abstract

Network optimization models have a wide variety of applications in operations research, management science, transportation, engineering design and other areas. Much has been done on the design of efficient algorithms for the solutions of very large problems. Difficulties remain, however, in building and implementing such models on the computer for practical use. Notwithstanding, recent developments in modelling languages, building network models remains the privilege of experienced modellers and is in general a time-consuming task. Expert systems and artificial intelligence methodologies could provide the link between practitioners and network optimization techniques. In this paper, we examine the role of expert systems in an operations research environment and identify the issues to be addressed in developing an expert system for network modelling. The discussion is illustrated with the development of a prototype expert system that solves matrix balancing problems using network optimization techniques. The system has the ability to identify the correct (nonlinear) network optimization model for each problem instance. A network problem is automatically set up and solved using a high-level modelling language. The results of the model and sensitivity analysis information are interpreted for the user by the expert system, in the vocabulary of the original description of the problem.