摘要
将决策者的心理偏好或权重理解为:"在总分相等的条件下,决策者认为各决策属性分数按其心理权重分布是最合理的."在此基础上,将通常的决策问题分解为求一系列(等总分)局部最满意解的子问题,然后再从所有局部满意解的集合中,找出全局最满意解(即(1)的解).并提出了一种直接面向决策者(经验性)心理标准及其评估变化过程的决策模型———属性坐标学习和分析的评估决策模型.
The optimal decision is the one whose evaluated scores about decision attributes are maximum amount all, and it is unique. However, it is not only difficult to be chosen from all decisions, but also maybe not exist. So usually we can only find out the most satisfied decision in many concrete problems. A kind of evaluation and decision approach based on analysis and learning of attribute coordinate system is presented in this paper. The key steps of the approach are as follows.1. A complex decision about overall situation can usually be separated into a series of simplex partial problems in which the most satisfied one can be find out by some special way respectively;2. The most satisfied decision at the lower sum of all levels, according to the mathematical interpretation of weight, could be considered as the decision whose distribution of scores is just consistent with the distribution of the psychology preference of a decision maker;3. At the moderate level, the most satisfied decision is selected from the samples which are given by the decision makers who can only choose from the certain level;4. If let the collection of satisfied decisions for whole be the union of the satisfied decisions of all levels, then it could be found out by the interpolation;5. In order to find out the most satisfied decision of from the whole collection of satisfied decisions, an identical satisfied function is also presented in this paper.
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2003年第2期182-188,共7页
Journal of Nanjing University（Natural Science）
基金
国家自然科学基金(60075016)
广西省自然科学基金(桂科自99010)
上海市高校科学发展基金(OG0104)
关键词
评估决策模型
机器学习
属性坐标分析法
定性映射函数
满意度
心理权重
评估标准
evaluation, decision, machine learning, analysis method of attribute coordinate, qualitative mapping, satisfying degree function
