【作者】 丁勇；

【导师】 梁昌勇；

【作者基本信息】 合肥工业大学 ， 管理科学与工程， 2011， 博士

【摘要】 在经济、社会和军事等诸多领域中，对一些复杂问题进行决策时，由于客观事物的复杂性及人类思维的模糊性，决策者在对诸如知识管理绩效、风险投资项目等进行方案评价时，难以用精确定量的形式进行判断，而运用“优”、“良”、“中”、“差”等自然语言形式进行判断，往往更加直观和方便。这类决策问题的特点是方案的属性值以自然语言形式给出，并且需要群体专家参与，称此类决策问题为语言型多属性群决策问题。因此，研究语言型多属性群决策问题具有重要的理论意义和实际应用价值。本文研究语言型多属性群决策问题，研究的主要内容如下：（1）研究一类属性值和属性权重信息均为已知且以语言评价信息形式给出的多属性群决策问题，提出了一种基于最小偏差和二元语义信息处理的主客观属性权重集成的决策分析方法。分析了基于最小偏差的属性权重求解方法的特点和优势，并给出了详细的决策分析步骤。通过算例分析中与一般方法的比较，说明该方法既能充分利用客观信息，又考虑到决策者的主观意愿，能够有效地减小决策群体意见的分歧，使决策结果更加合理。（2）针对一类属性权重信息完全未知的语言多属性决策问题，本文利用专家给出的语言决策矩阵信息，计算专家群体对方案属性值的判断偏差，从而确定属性权重。基于这种思路，给出求解属性权重的公式。最后利用二元语义加权平均算子计算得到各方案的综合评价值。（3）研究一类属性权重信息部分已知情况下的语言多属性决策问题，本文给出两种解决此类问题的方法。第一种方法是基于二元语义信息处理和TOPSIS的决策分析方法；第二种方法是基于线性规划模型和二元语义信息处理的多属性群决策方法。（4）针对一类属性值和属性权重均为不确定语言型，即语言区间值，专家权重为确定语言型的不确定语言多属性群决策问题，引入心态指标，提出一种基于心态指标的二元语义多属性群决策方法。该方法能够较好地反映在实际的决策过程中，不同决策者由于自身条件和外界环境的不同而具有不同的决策心态。（5）研究具有不同粒度语言评价信息的多属性群决策问题，分析目前常用的不同粒度语言信息的一致化方法。给出一种基于二元语义的不同粒度语言评价信息的一致化融合方法，并分析和证明其一些优良性质。针对一类属性值为多粒度不确定语言型、属性权重和专家权重均为确定语言型的多属性群决策问题，给出一种解决此类决策问题的方法和步骤，并通过案例分析说明方法的有效性和实用性。（6）探讨交互式群决策的基本原理及方法，分析一般交互式群体决策过程中存在的局限性，提出一种改进的交互式群体决策流程。引入交互式语言多属性群决策问题，设计了交互式语言多属性群决策一致性指标，分析了语言多属性群决策的交互过程，并给出交互式群决策过程中专家权重的调整方法。在对交互式语言多属性群决策问题描述的基础上，提出交互式语言多属性群决策模型与算法，最后通过案例分析说明本文方法的有效性。（7）分析企业知识管理绩效评价模型和方法的研究现状，在国内外学者相关研究成果的基础上建立企业知识管理绩效评价指标体系，运用论文中提出的语言多属性群决策方法对企业知识管理绩效进行评价，设计了企业知识管理绩效评价问题的决策流程，给出了详细的决策求解过程。通过案例说明本文方法在实际决策中的应用价值。更多还原

【Abstract】 In many domains such as society, economy and military, as the complexity of objective thingsand the ambiguity of human thinking, it is difficult to judge with precise quantitative form whendecision-makers evaluate the alternatives such as knowledge management performance and riskinvestment projects. But it is often more intuitional and convenient to evaluate alternatives using"excellent","good","medium" and "poor" and other forms of natural language. The characteristicof these decision problems is that attribute values of alternatives are given in the form of naturallanguage and that expert group participates in decision-making process. This decision makingproblem is called linguistic multiple attribute group decision making problem (LMAGDM).Research on such problem has important theoretical and practical value. This paper studieslinguistic multiple attribute group decision making problem, and the primary coverage is as follows:(1) A subjective and objective attribute weights integrated approach based on the minimizingdeviation and2-tuple linguistic information processing for multi-attribute group decision makingproblem with linguistic attribute values and attribute weights values are proposed. The paperanalyses the characteristics and advantages of attribute weights determining method based on theminimizing deviation. And the detailed step is presented. An example is given to illustrate that thismethod can make full use of objective information, and can take into account the subjective viewsof decision-makers, and can effectively reduce the differences of decision-making group opinion,and can make decision-making results more reasonable.(2) For linguistic multi-attribute group decision making problem with completely unknownattribute weights information, the deviation of attribute judgment of decision making group iscalculated according to linguistic decision matrix in order to determine attribute weights. Based onthis idea, the formula for solving the attribute weights is proposed. Then the comprehensiveappraisal value of each alternative is calculated by using the weighted averaging operator forlinguistic2-tuples.(3) The paper presents two solutions of linguistic multi-attribute group decision makingproblem with incomplete attribute weights information. The first one is decision analysis methodbased on2-tuple linguistic information processing and TOPSIS. The second one is a method basedon a linear programming model and2-tuple linguistic.(4) With respect to uncertain linguistic multi-attribute group decision making problem withuncertain linguistic attribute weights information, the2-tuple linguistic decision making methodbased on attitude indicator is proposed. This method can reflect that the different decision-makershave different decision-making mentality due to their own conditions and different externalenvironment in the actual decision-making process.(5) The paper analyses the common consistent methods of multi-granularity linguistic information fusion at present. The new consistent methods of multi-granularity linguisticinformation based on2-tuple linguistic is proposed. The paper analyses and proves its some goodproperties. On this basis, a decision analysis method and step is presented for multi-granularitylinguistic multi-attribute group decision making with uncertain linguistic information.(6) The paper discusses the basic principles and methods for interactive groupdecision-making, and analyzes the imitations of the general interactive group decision makingprocess. Based on this, a improved interactive group decision-making process is proposed. Withrespect to interactive linguistic multi-attribute group decision making problem, the consistencyindicators are designed. The paper analyses interactive process, and presents expert weightadjustment method in the interactive linguistic multi-attribute group decision making process.Based on this problem description, the model and algorithm for interactive linguistic multi-attributegroup decision making is proposed. Finally, an example is given to illustrate that this method iseffective.(7) The paper analyses research status of the model and method of enterprise knowledgemanagement performance evaluation. The index system of enterprise knowledge managementperformance evaluation is constructed based on the related research of scholars at home and abroad.Applying the LMAGDM method proposed by this paper to evaluate enterprise knowledgemanagement performance. The decision-making flow of enterprise knowledge managementperformance evaluation problem is designed, and the detailed solving process of decision-making isgiven. Finally, an example is given to illustrate the application value of this method.更多还原