MIDAS: An Expert Debt Management Advisory System

Summary

This paper describes an ongoing university/industry cooperative project within the context of the Courseware Development Project at Dalhousie University. The project integrates artificial intelligence and management science approaches within an expert decision support system to assist the treasurer of a public electric power utility in his debt management role. The prototype system under development provides optimization, simulation and object-oriented database components at the command of a menu-driven user interface and an interactive rule base. The rule base manages model runs and interprets model results to the user through text, numeric and graphic outputs. Scenario-oriented stochastic programming suggests optimal flotation and refunding decisions; the resulting suggestions can be refined heuristically through the rule base and explored in more detail with stochastic simulations consistent with the postulated scenarios. The system is being written in KEE, LISP and FORTRAN and runs in a distributed hardware and software environment on an Ethernetted Unisys Explorer and DEC MicroVAX II. It represents the use of expert system methodology for advanced mathematical model management, explanation and user interfacing in a rich—but specific—application domain.

This research and the Dalhousie Courseware Development Project are supported by the Bank of Nova Scotia, Digital Equipment Corporation, Unisys Corporation, the Nova Scotia Power Corporation, the Natural Sciences and Engineering Research Council of Canada and the Social Sciences and Humanities Research Council of Canada