【作者】 骆公志；

【导师】 杨晓江；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2009， 博士

【摘要】 数量方法是高等教育评估方法中最为重要的方法之一。由于评估环境的复杂性与动态性,使得评估主体所面临的数据具有多样性与不确定性,传统的评估技术已不能适应高等教育评估的现实要求。本文研究了基于粗糙集理论的高等教育评估方法,针对不完备信息条件下的基于拓展优势关系的粗糙决策分析方法,提出了限制优势关系、确定优势关系、限制相似优势关系等新的拓展优势关系,对原有的拓展优势关系进行了有效的改进和推广;针对属性约简结果的不唯一性,以条件信息熵为工具,给出了属性约简的择优方法;研究了属性约简的算法,提出了一种基于论域划分类归并的粗糙集属性约简新的算法;在以上研究的基础上,对江苏省高职高专人才培养水平评估工作作了实证分析。本文的主要创新点有:(1)在不完备信息决策系统中,提出了限制优势关系以及确定优势关系的概念,更合理地处理了属性未知值与属性已知值的比较问题。对限制优势关系以及确定优势关系分别给出了求得相应粗糙集属性约简的具体方法。对比结果表明,使用新的拓展优势关系可以获得可信度更高的粗糙决策规则。(2)将确定优势关系的概念推广到粗糙模糊集中,并在粗糙模糊集中提出了限制相似优势关系。在确定优势关系及限制相似优势关系粗糙模糊集的基础上,提出了属性约简的概念,给出了求得这些约简的具体操作方法。使用基于确定优势关系以及限制相似优势关系的粗糙模糊决策分析方法,可以获取不完备模糊目标信息系统的“至少”、“至多”决策规则。(3)为排除决策过程中噪声因素的干扰,基于确定优势关系提出了变精度粗糙决策分析方法。以从多个粗糙集属性约简中选择最优的约简为目的,分析了作为度量工具的现有条件信息熵在应用过程中的缺陷。借鉴变精度粗糙集理论的思想,在对阈值参数进行二次选择的基础上,提出一种新的条件信息熵。基于新的条件信息熵设计了一种变精度优势粗糙集属性约简的择优算法,克服了现有条件信息熵的不足。(4)针对现有求属性约简算法的不足,以论域的不同划分与条件属性的不同子集确定的等价关系间的一一对应关系为出发点,研究了粗糙集属性约简与论域划分类归并之间的内在联系,利用粗糙集模型某些属性约简的正区域不变性,提出一种基于论域划分类归并的粗糙集属性约简新算法。(5)从决策理论的角度出发,在以上粗糙集理论研究基础上,对江苏省高职高专人才培养工作水平评估数据库进行实证分析,揭示了江苏省高职高专人才培养工作水平评估一级指标与院校评估整体结论之间的内在规律,为教育主管部门、评估机构、参评院校提供一些有价值的决策信息及政策建议。更多还原

【Abstract】 Quantitative method is one of the most important methods of higher edueation evaluation. Because of the complexity and dynamic of the environment, the evaluation subject has to deal with diverse and uncertain data, and the traditional methods could not meet the practical requirements of the higher education evaluation. This paper mainly discusses methods of higher edueation evaluation based on rough set theory. New expansion of dominance relations such as limited dominance relation, definitive dominance relation and limited similarity dominance relation are proposed for multi-attribute decision making problems with preference and incomplete information. For the results of attribute reduction are not unique, a preferred method in which conditional information entropy is used to measure reduction of attributes is proposed. Intrinsic relationship between attributes reduction relation of rough sets and classification of universe is researched and an efficient attribute reduction algorithm of rough sets based on classification merging is given by using an invariance of positive region of some rough sets. Based on the research on rough set theory, the empirical analysis of the talents cultivation level assessment project for higher technical and vocational institution of Jiangsu province are made. The main innovations of the paper are as follows:(1) In the incomplete information decision-making system, the limited dominance relation and definitive dominance relation proposed are more rationally to deal with the known and unknown attribute values. Based on the limited dominance-based and definitive dominance-based rough sets, two types of knowledge reductions are proposed. Then, the practical approaches to compute the reductions are presented. One can obtain higher quality of decision rules by using the limited dominance-based and definitive dominance-based rough sets model.(2) The concept of definitive dominance relation and limited similarity dominance relation are proposed in the rough fuzzy set. And based on above new extended dominance relations, two types of knowledge reductions are proposed. Then, the practical approaches to compute the reductions are presented.“At least”and“at most”rules can be obtained by using definitive dominance-based and limited similarity dominance-based rough sets model in the incomplete fuzzy objective information system.(3) In order to exclude the interference of the noise factor, based on definitive dominance relation, an analysis method of decision-making based on variable precision rough set is proposed. Then, the disadvantages of the current conditional information entropy are analyzed when conditional information entropy is used to measure reduction of attributes of dominance-based rough sets. Variable precision rough set theory is used as a basis for proposing a new condition information, with another parameter choosed. Then, a new algorithm based on new condition information entropy is designed to choose reduction of attributes of Variable precision dominance rough set, which is better than others.(4) Intrinsic relationship between attributes reduction relation of rough set and classification of universe is researched through one to one mapping relation between classification of universe and equivalence relation determined by subset of attribute set. Using an invariance of positive region of some rough sets, an efficient attribute reduction algorithm of rough set based on classification merging is proposed.(5) Based on the decision-making theory and the research on rough set theory, the empirical analysis of the talents cultivation level assessment project for higher technical and vocational institution of Jiangsu province are made. The Inherent discipline between the class 1 index and the general evaluation is revealed which can provide some valuable decision-making information and policy recommendations for educational authorities, evaluation agency and assessed institutions.更多还原