An intelligent decision support system for debt management

In order to be used effectively by managers, decision support systems (DSS's) which use complex models require the assistance of human intermediaries in formulating, applying and managing models and in explaining and interpreting their output. Expert systems techniques offer ways of incorporating intermediary functions into DSS's so that they can be used directly by managers who are not modelling experts.;MIDAS' design is based on five guiding principles: frame-based knowledge representation, object-oriented modelling, spreadsheet-oriented financial model structures, separation of knowledge and control or reasoning, and integration of models and heuristics through an underlying domain representation. A prototype system has been implemented which incorporates these principles for simulation, user assistance in task selection, and output management. Detailed design descriptions and documentation are presented in the dissertation. The research demonstrates the feasibility of (a) hierarchical debt planning supported by multiple complex models and heuristics and (b) knowledge-based decision support for debt management. It also suggests a number of further research questions related to intelligent model management and explanation.;This dissertation reports on a project to integrate expert system and financial modelling techniques in a DSS for corporate debt planning. The system, called MIDAS (Manager's Intelligent Debt Advisory System), supports hierarchical planning in which a stochastic linear optimization model suggests a borrowing plan based on the user's problem description, the plan is modified using heuristics incorporated into the system, and stochastic simulation and (deterministic) cash flow projection are used to test the detailed implications of the resulting plan alternatives. The system is designed to assist the user in task selection and execution; model formulation, solution and modification; result analysis; and result explanation and interpretation.