决策系统中基于数据的属性交融性度量理论与方法

In background of practical applications, such as facility location and performance evaluation, the project focuses on the multi-attributes decision making problems under fuzzy environment, in which the attributes are interdependent and interrelated. We first explore the features and limitations of existing methods. Then according to the change of knowledge carrier induced by that of attributes set in the decision information system, we systematically study the metric of attributes interaction based on data and further study the multi-attribute decision making theories and methods. The work mainly includes: 1) by different knowledge acquisition methods, we discuss the measurable description system based on basic knowledge factors, which are denoted by the core samples, rules, fuzzy number, respectively; 2) we study the metric of attributes correlation based on the changes of knowledge, and further construct the determination mechanism of attributes interaction, which not only reflects the importance measures, but satisfies fuzzy measure structural characteristics; 3) by taking the fuzzy integral as synthesis operator, we study the multi-attribute decision making methods, and further construct several operable decision making models with structural characteristics; 4) we build the description system for consistency and redundancy of complex datasets based on the interaction metric; 5) finally, we analyze the features and effectiveness of the models from different aspects with case study. Accordingly we further develop the attributes interaction metric system and multi-attributes decision system based on the datasets.

本项目以资源分配、绩效评估等问题为实际背景和应用面向，围绕模糊环境下具有交互作用的多属性决策问题，在分析现有方法的特点与不足基础上，以决策数据集为基础、以属性集的变化导致的知识的变化为依据、采用多方法集成与融合的策略，系统地研究基于数据的属性交融性度量和多属性决策理论与方法，主要内容包括：1）针对不同的知识获取方法，分别研究以核心示例、规则和模糊数为基本知识因子的知识的可量化描述体系；2）研究以知识的变化为基础的属性关联性度量方法，构建反映属性重要性且满足模糊测度结构特征的属性交融性度量机制；3）系统地研究以模糊积分为综合算子的、基于数据的多属性决策方法，构建几种具有结构特征和可操作性的具体决策模型；4）构建基于交融性度量的复杂数据的一致性和冗余性刻划体系；5）结合具体案例，从不同的角度分析和验证所建模型的特点和有效性，有针对性地开发基于数据的属性交融性度量系统和多属性决策系统。

本项目采用学科交叉的手段，围绕模糊环境下具有交互作用的多属性决策问题，针对属性重要性度量问题进行了较为系统的讨论。1) 以隐藏在决策信息系统中的知识为载体，以属性集的变化所引起的知识的变化为依据，讨论了正域、决策类的上（下）近似与系统中知识的关联特征，给出了几类满足模糊测度结构特征的基于知识变化率的属性重要性度量方法，并讨论了其结构特征和构建策略，进一步提出了基于综合知识变化率的属性约简算法；2) 以数据系统为基础、以隐藏在数据系统中的知识为载体、以集合间的包含程度为依据，提出了一种基于知识可靠性的数据删除方法，讨论了知识因子在子数据系统中的变化规律，建立了基于数据效用的属性重要性度量；进而，通过理论证明和实例计算讨论了其取值规律和结构特征，并进一步提出了相应的属性约简算法。本项目已完成计划任务书中的研究内容，在《Information Sciences》、《Applied Mathematical Modelling》、《Knowledge-Based Systems》、《Enterprise Information Systems》、《Systems Research and Behavioral Science》、《International Journal of Machine Learning and Cybernetics》、《ICIC Express Letters 》等国内外重要学术期刊和国际会议上发表相关学术论文33篇，其中有10篇被SCI收录、23篇被EI收录。培养博士研究生2名，硕士生9名，青年教师2 名。