【作者】 王建明；

【导师】 陈守煜； 汪德虎；

【作者基本信息】 大连理工大学 ， 水利水电工程， 2004， 博士

【摘要】 复杂系统决策理论是当今系统科学研究的前沿领域，随机性和模糊性是复杂系统不确定的两大主要表现形式，模糊集理论的发展为决策科学提供了重要的理论与方法。本文首先概括分析了复杂系统优化决策研究现状，简要介绍了陈守煜教授创立的工程模糊集理论优化决策的基本原理和发展，以复杂水资源系统优化决策为主要研究背景，以工程模糊集理论为基础，就复杂系统多目标模糊识别优化决策理论与应用进行探讨，取得的主要研究成果如下： (1)．在陈守煜教授提出的模糊聚类循环迭代模型的基础上，提出了一种考虑聚类目标在不同类别具有不同目标权重的循环迭代模糊聚类算法，避免了模糊C均值聚类算法引入模糊加权指数带来的聚类不确定性，其聚类结果是超球的，克服了欧氏距离只能聚类球状数据的缺陷，对多种分布类型的数据均有良好的聚类效果；提出样本目标值残缺情况下的三种等效聚类算法，可使残缺数据样本有效地参加聚类，减少样本数据残缺对聚类的影响；提出一种将已知样本的先验知识融合到模糊聚类过程中的半监督循环迭代聚类模型，较为有效地克服模糊聚类为无监督模糊识别存在的弱点。 (2)．半结构性决策是多目标决策的难点，在工程模糊集单元系统决策理论基础上，认为决策者给出二元比较判断往往是不一致的，且对作出每个判断的把握程度也是不同的，提出在两阶段构造偏好关系矩阵基础上，将决策者对每个判断的把握程度作为可信度，以判断偏差最小平方法确定目标权重和方案优越度；在二元比较残缺可接受范围内，该模型可有效地处理判断信息残缺等情况的目标定权和方案优越度处理，给出了简易判断残缺模糊偏好矩阵为可接受的条件；建立一种将主、客观赋权综合的决策方案模糊识别模型，有效地融合主客观赋权信息，提高决策目标定权的准确性。 (3)．在陈守煜教授提出的模糊模式识别模型基础上建立了决策信息不完全确知的多目标决策集成模型。该模型可综合处理权重信息不完全、分级标准不完全、方案集评价不完全、方案集目标不完全等决策信息不完全确知情况。丰富了模糊聚类、模糊模式识别、模糊决策统一理论。示例分析表明该模型具有物理概念明确、简便易用的特点。 (4)．流域水资源丰富度是制定可持续发展规划的一项重要指标，在水资源丰富度评价中目标权重的确定是关键问题之一。在模糊模式识别模型的基础上，以专家经验、知识为监督，提出一种可同时确定评价目标权重和流域水资源丰富度的监督模糊模式识别模型，在辽西沿海9个流域水资源丰富度评价中取得较为满意的结果。 (5)．群决策可以集结决策群体的智慧，是解决复杂系统决策问题的有效方法。如何有效地将决策成员偏好模式结集成反映决策群体意愿的模式一直是群决策研究的热点。在工程模糊集理论基础上，提出了三个层次进行群决策集结方法:加权平均法、群决策权最小平方法和群决策监督模糊模式识别模型，分别从决策者二元比较判断平均、决策者二元比较判断综合最优到考虑决策群体主、客观权重对决策方案进行综合模糊模式识别进行优化决策，以适应不同的决策环境。 (6).多目标动态规划经常涉及到定性、定t目标，在陈守煌教授提出模糊优选动态规划的基础上，对定性、定量目标统一的相对优属度进行了进一步探讨，提出了适合定性目标递推的定性目标二元比较递推法和各阶段决策方案分组综合确定法两种定性目标相对优属度确定方法，将主、客观权重综合的模糊模式识别与动态规划模型结合，拓展了陈守煌教授提出的动态规划级别特征值法和阶段递推法。最后对全文作了总结，并对有待进一步研究的问题作了分析和展望。更多还原

【Abstract】 Complex systems decision theory is one of the heading fields in the research of system science, and randomness and fuzziness are the main behaving forms of uncertainties existed in complex systems. Fuzzy set theory offers crucial theory and methodology for decision making science. The research status of complex systems decision making is generalized first, then the principle and development of engineering fuzzy set theory proposed by Professor Chen Shouyu is briefly reviewed. With the background of complex water resources systems optimization and decision and on the basis of engineering fuzzy set theory, multiobjective fuzzy recognition, optimization and decision theory for complex systems and applications are investigated, and the main research results are listed, as followed.(1) Founded upon the cross iterative clustering fuzzy algorithm (CIFCA) proposed by Professor Chen Shouyu, a CIFCA algorithm, with the consideration of the weights of the objects to clustered attributed to different clusters, is presented first, which tries to avoid the clustering uncertainties produced by the fuzzy weighted parameter in the FCM algorithm. With hyperspherical clustering results, it can partition different distribution type data with fine clustering results. Then, three equivalent clustering algorithms are provided for the dataset with fragmentary samples, which decreases the influence of incomplete data on clustering. At last, a semi-supervised cross iterative fuzzy clustering algorithm, with the integration of transcendental knowledge into clustering, is proposed, which can effectively handle the weakness of unsupervised fuzzy recognition.(2) Semi-structural decision making is a difficulty of multiobjective decision making. The pairwise comparison provided by the decision makers is frequently inconsistent, and the reliability of each judgment is different. Based on engineering fuzzy set element system decision making theory and the preference matrix constructed through two steps, the objective weights and schema’s superiority are determined via the method of the least squared errors with the consideration of the reliability of decision makers’ judgment. Within the acceptable range of incomplete pairwise comparison, the proposed model can specify the objective weights and schema’s superiority under the condition of incomplete judgment information, and the conditions, which can simply judge incomplete fuzzy preference matrix acceptable or not, are given. A fuzzy recognition model is constructed, which can effectively integrate subjective and objective weights and promote the precision of the weight assessment.(3) On the basis of the fuzzy pattern recognition model proposed by Professor ChenShouyu,, a integrated multiobjective decision making model with incomplete decision making information is established, which can handle different kinds of decision making environment, such as incomplete weights, incomplete ranking standards incomplete judgment of the scheme set, incomplete objectives of the scheme set, etc. Experimental analysis shows that this model has a feature of clear physical concept and simple application.(4) The abundance degree of basin water resources is a significant index in the establishment of sustainable development planning, and the weight assessment is one of the key problems in the evaluation of the abundance degree of water resources. Grounded upon the fuzzy pattern recognition model, under the supervision of experts’ experience and knowledge, a supervised fuzzy pattern recognition model, which can simultaneously determine the objective weights and the abundance degree of basin water resources, is presented, and has a satisfactory result in the evaluation of the abundance degree of Liaoxi basin water resources.(5) With the intelligence of a group of decision makers, group decision making is a effective method for complex systems decision making problems. The method for effectively integrating the preference of each decision maker into the group preference is a hotspot in the research of gro更多还原