【作者】 胡东滨；

【导师】 陈晓红；

【作者基本信息】 中南大学 ， 管理科学与工程， 2008， 博士

【摘要】 问题管理系统(PMS)是决策支持系统的重要子系统,更是问题驱动型决策支持系统的核心子系统,其中,决策问题的分析处理与求解等问题管理的方法以及问题管理系统的开发,一直就是该领域研究的重点和难点。尽管如此,在面向人机交互的决策问题管理方法方面仍然存在很多难题没有很好得到解决,包括决策问题的形式化描述,用自然语言表述的决策问题的识别与理解,决策问题的子问题分解,决策问题求解方法的智能匹配及运用,决策问题库的有效组织等等。本文针对这些问题,通过分析决策问题的普遍特征,在综合人与计算机对决策问题处理的各自优势基础上,探讨更具智能性的决策问题管理和分析的一般方法,并以此研制出问题管理系统开发组件,为问题驱动的决策支持系统的开发提供具有较强实用性的手段和工具。本文的主要研究内容及研究成果包括:(1)决策问题的特征分析与描述方法的研究。将决策问题看成一种复杂的系统,在分析决策问题系统的内部结构及特征的基础上,提出一种面向对象的问题向量空间模型,用以作为决策问题系统的描述方法。同时,运用问题向量空间模型研究探讨决策问题属性类型、属性结构和问题理解与求解过程中的属性取值状态的变化规律,以及问题属性识别与分析同问题求解及其过程中子问题分解的关系,得到从决策问题描述,到问题分析理解、子问题产生,以及问题求解方案形成的问题处理一般方法。(2)决策问题智能处理方法的研究。在根据决策问题特征分析而得到问题处理的一般方法的基础上,将知识管理技术引入到这一处理方法中,研究基于知识管理的决策问题智能处理方法所涉及的三个核心问题:问题管理系统的问题库组织方式、问题的知识结构以及问题智能理解与求解过程和方法。并以此提出一种嵌入到知识库的问题库管理方法,研究了问题库嵌入到知识库的方式、相应的问题库及知识库的基本结构。同时,根据决策问题处理所涉及的各种知识的组成和数据结构,提出了问题的知识结构.另外,给出了根据问题的描述识别问题属性、根据问题属性识别问题类型和求解方法、以及根据问题属性的取值状态来识别子问题等基于知识管理的问题理解与求解的具体流程和机制。(3)决策问题智能识别相关模型与方法的研究。在上述决策问题智能处理方法的基础上,建立问题属性识别、问题类型识别以及子问题识别的模型和算法。通过给出中文语句分词算法的选择依据,建立了基于分词算法、子句分割算法和句法分析算法的决策问题表层属性智能识别模型;建立了基于双关键词搜索规则和问题类型知识库匹配规则的决策问题隐含属性智能识别模型;建立了基于子问题产生规则和构建规则的子问题智能识别模型。(4)决策问题管理系统开发组件研究。依据本文对决策问题特征的研究和所提出的问题智能处理方法与模型,设计出通用问题管理系统的功能结构。将该功能结构组件化,设计出问题管理系统开发组件(PMS-DC),详细探讨了PMS-DC中所需包含的对象、类和控件及其属性、方法和事件,并进一步研究了PMS-DC内部各种类与控件的相互关系,以及与互联网环境下通用决策支持系统生成器(i-GDSSG)中其他组件的关系,设计了PMS-DC的外部接口,给出了在专用DSS中的组件部署策略。(5)组件的实际应用。运用本文设计和开发的PMS-DC及i-GDSSG的相关组件,实际开发了一个问题驱动型DSS,即中国固体矿产业持续技术创新支撑体系决策支持系统(CMTIDSS)。应用效果表明,采用PMS-DC及相关组件开发面向问题的专用DSS,不仅可提高开发效率,且能有效地管理和处理中文自然语言表述的决策问题。更多还原

【Abstract】 Problem Management System (PMS) is the important subsystem of DSS and the core subsystem of problem-driven DSS. Problem management methods (e.g. analyzing problem and problem solving) and development of problem management system are the difficulty and emphasis in this research area. A lot of difficult problems about human-machine communication system faced decision problem management system remained unsolved including formal description of decision problem, identification and understanding of decision problem expressed in natural language, sub-problem decomposition of decision problem, intelligent matching and application of solving method for decision problem, effective organization of decision problem base and so on. With respect to these problems, this dissertation discusses the more intelligent general method of management and analysis for decision problem by analyzing the common characteristic and integrating each advantage of human and computer, and researches development component of PMS to provide more practical means and tools for the development of problem driven DSS.The main research content and result of this dissertation include:(1) Research on characteristic analysis and description method of decision problem. Decision problem is regarded as a complex system. On the basis of analyzing interior structure and characteristic of decision problem, an object-oriented problem vector space model is proposed to act as description method of decision problem system. At the same time, problem vector space model is applied to research the attribute type, attribute structure, the change law of attribute valuing state in the course of problem understanding and solving, and the relation of sub-problem decomposition among problem attribute identification, analysis and problem solving. Then general method of problem handing is drawn from description, analysis and understanding, sub-problem causing and problem soling scheme forming of decision problem.(2) Research on intelligent method of decision problem management. On the basis of general method of problem solving acquired from characteristic analysis of decision problem, knowledge management technology is introduced to the solving method to research three core issues involved in knowledge management based intelligent handing approach of decision problem: problem base organization mode of PMS, problem knowledge structure, problem intelligent understanding, and problem intelligent solving process and approach. A problem base management approach embed in knowledge base is suggested to research the mode of problem base embed into knowledge base, basic structure of corresponding problem base and knowledge base. Meanwhile, problem knowledge structure is put forward according to varieties of knowledge composition and data structure involved in decision problem handing. In addition, concrete flow and mechanism of knowledge management based problem understanding and solving is presented according to problem description identification type and solving method, valuing state of problem attribute for identifying sub-problem.(3) Research on relevant model and technology of decision problem intelligent identification. On the basis of above-mentioned intelligent handing method of decision problem, problem attribute identification, problem type identification, model and algorithm of sub-problem identification are built. By supplying choosing cause of Chinese sentence segmentation algorithm, intelligent identification model of decision problem surface layer attribute is built based on segmentation algorithm, clause division algorithm and syntax analysis algorithm. Intelligent identification model of decision problem concealed attribute is built based on double key words searching rule and problem type knowledge base matching rule. Intelligent identification model of sub-problem is also built based on sub-problem producing rule and constructing rule.(4) Research on development component of decision problem management system. Function structure of general problem management system is designed according to the characteristic of decision problem, proposed problem intelligent handing method and model. The function structure is moduled to design out problem management system development component (PMS-DC). The object, class, control, its attribute, method and event contained in PMS-DC are discussed in detail. The relation among variety of classes and controls in PMS-DC is discussed. And the relation among other components in general DSS generator under internet environment (i-GDSSG) is also discussed. Exterior interface of PMS-DC is designed and component distributing strategy in special DSS is present. (5) Practical application of component. The designed and developed PMS-DC and relevant component of i-GDSS in this dissertation is actually a problem driven DSS named CMTIDSS. Applying effect shows that developing the problem-oriented special DSS by introducing PMS-DC and relevant component not only improves developing efficiency, but also effectively manages and deals with decision problem expressed in Chinese natural language.更多还原