一种交互式多目标优化方法

An Interactive Multiobjective Optimization Method

本文从实用的角度提出一种新的交互式多目标优化方法——轮流打击目标法。该法要求决策者在每步迭代中指出在各现行目标值中他认为哪个目标值相对最大.此外,该法每步迭代所要求解的单目标优化问题只包括两个目标函数。所以,该法的计算是很简单的,而且谈法还具有决策容易,综合面广,折中性强,对目标个数不敏感的优点。它是一种有效的、实用的多目标优化方法。

For practical purposes,the paper proposes a new method for solving mul- tiobjective optimization (MOO) proclems-the Striking Objective in Turn (SOIT) method.It is interactive.At each interation,the decision maker is required to point out which of the current objectives values,he thinks,is the relatively maximal value.Only two objective functions are considered in the single- objective optimization problems that should be solved at each iteration.So the computation of SOIT is very simple.Moreover,SOIT method possesses the advantages of making decisions easily,synthesizing widely,compromising appropriately,not being sensitive to the number of objectives and so on.It is a practical and effective MOO method.