多目标优化决策方法的研究方问探讨——走向智能化

本文对多目标优化决策(MODM)中如何逐步引入知识工程手段、专家系统技术进行研究和探索。文中结合MODM的三类求解方法分别论述了:非劣解生成方法中的知识利用与控制;将事先偏好结构视作一个智能性的结构;逐步结合偏好方法的求解依赖于决策者的经验和判断;基于知识的专家系统(ES)具有求解MODM中带有不良结构问题的特征。文中最后提出ES与模型相结合形成的智能化MODM它将是今后发展的研究方向。

On the Artificial Intelligence Approach to the Multiobjective Decision Making Methods

Since the seventies, many research papers on Multiobjective Decision Mak-ing (MODM) have centered around three MODM methods: (1) Generation ofNoninferior Solutions, (2) Priori Preference, and (3) Progressive Articulationof Preference. These three methods all suffer from the following shortcoming:in the process of finding solutions, some information is either neglected orbiased by personal preference. The elimination of such a shortcoming is the aimof this paper. In this paper, the information required for solving problems is processedinto intelligence information by means of knowledge engineering. This paperemphasizes the role of knowledge and experience. Through formalization, thespecialized knowledge of a particular field--including both the relevant theore-tical knowledge and the viewpoints and judgements of experts--is stored in aknowledge base. Combining MODM with such an established knowledge systemmakes problem solutions more effective. Even ill-structured problems, unsolva-ble with conventional three MODM methods, can be dealt with. Furthermore,the decision-making process becomes more transparent and flexible than that ofthe conventional three MODM methods. The above is essentially a significant improvement on the conventional me-thod of priori preference: establishing an intelligence system to assign priorityof objectives and treating priori preference structure as an intelligence structure.It is believed that the author is the first to propose such an improvement. The above improvement has proved preliminarily successful when appliedto decision making and planning in business management. By applying knowledge engineering to MODM, this paper indicates it ispossible to introduce artificial intelligence into MODM research.