模型管理系统中问题求解策略的研究

A Problem-Solving Strategy for Model Management Systems

本文将半结构化或非结构化问题的求解策略与模型单元的管理方式有机地结合起来,提出了基于不精确推理的问题求解策略以及相应的模型描述与操作机制。文中还提出了基于问题求解目标树型分解的问题结构化方法。

The model management system(MMS)has been one of the central issues in the decision support system(DSS). Three drawbacks are found in the currently used model management systems:1)The decision-maker's intuition, experience and preference have not yet been incorporated into model representation and manipulation; 2)The uncertainties (such as fuzziness, probabilities etc.)in the decision environment have been overlooked in model representation; 3) There lacks optimization-based inference mechanism for model manipulation. In this paper, an inexact-reasoning approach for model-based problem solving and the corresponding scheme for model representation and manipulation is proposed. The primary concern is to develop a new technology for model management in uncertain situations, and combine the process of problem-solving with that of model management, while the decision-maker's intuition, experence and judgment will be used to guide the processes of problem-structuring and solving. An heuristic search algorithm for model selection is also presented.