一类偏好解聚的非线性多准则分类决策方法研究

发布时间：2018-03-03 01:18

本文选题：多准则分类决策　切入点：偏好解聚　出处：《中南大学》2004年硕士论文　论文类型：学位论文

【摘要】：多准则分类决策是多准则决策的一个重要研究分支,是管理决策以及社会生活中常见的一类问题。简单来说,它处理的是多个方案在多个准则下如何分门别类的问题。在实际决策中,方案的偏好值并不一定是各准则偏好的线性组合,即准则间并非完全独立而是相关联的,同时决策者也很难预先确定出所有偏好函数和权系数,通常只能给出相应的不完全信息。由此,本文针对这类不完全信息下非线性情况的多准则决策问题展开了系统研究。本文首先在查阅大量文献的基础上,论述了国内外多准则决策和分类方法的研究现状,分析了目前已有方法,如MAUT、ELECTRE/PROMETHEE、UTA/UTADIS、M.H.DIS的局限性或不足。其次,改进并构建了具有非线性偏好聚合函数的多准则分类优化模型,并利用偏好解聚和层层递进的思想,进一步给出了一种基于偏好解聚的多准则层次分类决策方法,该方法集中了现有PROMETHEE、ELECTRE和M.H.DIS方法的优点,克服了其中的不足。对分类优化模型的求解是本文研究的一个重点和难点。在分析了常规方法失效而进化算法具有求解优势的情况下,引入进化策略,并结合动态及模拟退火罚函数构造出了一种较好的求解算法,通过计算机编程实现,有效地解决了模型计算难度大、求解时间长的问题。然后基于此算法,设计并开发了以上述模型为核心的交互式多准则分类决策系统。该系统运算速度快,可靠性高,并具有一种交互反馈机制,使决策者可以监控中间过程,修正结果,逐步降低人的主观性影响,从而快速获得有效的分类结果。最后通过算例对比分析和验证了此方法的有效性和科学性,为其它相关领域和学科中的类似决策问题提供了一种有益参考。
[Abstract]:Multi-criteria classification decision making is an important branch of multi-criteria decision making. It is a common problem in management decision making and social life. It deals with the problem of how to classify multiple schemes under multiple criteria. In practical decision making, the preference value of the scheme is not necessarily a linear combination of the preferences of each criterion, that is, the criteria are not completely independent, but are related to each other. At the same time, it is difficult for decision-makers to determine all preference functions and weight coefficients in advance, and usually can only give the corresponding incomplete information. Therefore, a systematic study is carried out on the multi-criteria decision making problem in nonlinear cases with this kind of incomplete information. Based on a large number of references, this paper first discusses the current situation of research on multi-criteria decision making and classification methods at home and abroad, and analyzes the limitations or shortcomings of the existing methods, such as MAUTTECTRER / UTADISM / UTADISM / H.DIS. The optimization model of multi-criteria classification with nonlinear preference aggregation function is improved and constructed, and a multi-criteria hierarchical classification decision method based on preference clustering is presented by using the idea of preference deconcentration and hierarchical progression. This method focuses on the advantages of the existing Prometheek-ELECTRE and M.H.DIS methods, and overcomes the shortcomings of them. It is an important and difficult point to solve the classification optimization model in this paper. After analyzing the failure of conventional methods and the advantage of evolutionary algorithm, the evolutionary strategy is introduced. Combined with dynamic and simulated annealing penalty function, a better algorithm is constructed. Through computer programming, the problem of difficult model calculation and long solution time is solved effectively. An interactive multi-criteria classification decision system based on the above model is designed and developed. The system has the advantages of fast operation, high reliability, and an interactive feedback mechanism, which enables the decision makers to monitor the intermediate process and correct the results. Gradually reduce the subjective impact of human, so as to quickly obtain effective classification results. Finally, the validity and scientificalness of the method are verified by comparison and analysis of numerical examples, which provides a useful reference for similar decision making problems in other related fields and disciplines.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2004
【分类号】：C934

【参考文献】