【作者】 李鸿；

【导师】 胡学钢；

【作者基本信息】 合肥工业大学 ， 计算机应用技术， 2006， 硕士

【摘要】 本文在基于粗糙集理论的基础上,探讨知识粗糙性的表达方法和应用等方面问题的研究。 在知识处理中,知识被定义为一个或多个信息之间的关联(关系或联系)。知识是人类实践经验的总结和提炼,具有抽象和普遍的特性。人类在认识事物时总是将事物的全体进行抽象,即进行信息的颗粒化,在这一过程中,尽管信息个体的某些细节丢失了,但颗粒化的信息更有意义,更容易理解。在对颗粒化的信息进行解释和理解时,会产生某种不精确性,把这种不精确性称为知识的粗糙性。研究知识粗糙性,不仅具有重要的理论上、方法论上的意义,而且具有重要的实际应用价值。 粗糙集(Rough SetS)理论是由Pawlal(教授于20世纪80年代初提出的一种用于处理不确定性和含糊性知识的数学工具,其基本思想是:基于分类的观点,在近似空间(知识库)中研究如何在保持分类能力不变的前提下,通过知识约简,导出概念的分类规则,然后将此规则用于未知数据,以决定如何作出决策。它无需提供相关数据集合外的任何先验信息,适合于发现数据中隐含的、潜在有用的规律,即知识,找出其内部数据的关联关系和特征。近年来,基于粗糙集的知识粗糙性研究和应用取得了很大的成功,已成为软计算方法的重要分支,其涉及的领域包括模式识别、机器学习、决策分析和决策支持、知识获取、知识发现等。因此,粗糙集理论为研究知识粗糙性的研究提供了框架。 本文阐述了基于粗糙集理论的知识粗糙性研究背景、意义和现状;总结了粗糙集理论的基本思想;分析了基于粗糙集理论的知识粗糙性的产生、性质及实质;讨论了知识粗糙性的度量问题、结构问题、表示问题和应用问题;说明了知识粗糙性的研究对人类认识问题和解决问题有一定的推动作用。在上述研究基础上,论文获得了以下成果:(1)在论述知识粗糙熵、知识粒度等概念的基础上,首次提出了知识粗糙性的互粗糙信息这个概念;(2)给出了基于粗糙集理论的知识粗糙性的两种表示形式,即粗糙熵表示和粒度表示;(3)论述了知识粗糙性与粗糙熵、粒度的一些关系,并分别建立了知识粗糙性与粗糙熵和粒度之间的单调关系;(4)基于两种表示形式,分别给出了知识表示系统的属性简化的基本过程,分别设计了知识约简算法,通过实例分析,表明这两种算法是有效的。更多还原

【Abstract】 The present thesis attempts to investigate knowledge roughness research based on Rough Set Theory.In knowledge processing, knowledge by definition is one or more information connections (relations or relation). Knowledge is to summarize and abstract human experience, characterized with abstraction and universality. While human understanding an object, the object is generalized and abstracted, namely the information is granularized. Although individual details of information lost, the granularized information is more significant, easier to understand. During understanding and explanation of the granularized information, a certain of inaccuracy which will be produced is called the knowledge roughness. Research of knowledge roughness is important in theory, and methodology, moreover, valuable in application.Rough Set Theory, introduced by Professor Pawlak in the early 1980’s, is a mathematical tool to deal with vagueness and uncertainty. Its basic idea goes as follows: Based on viewpoint of classification, classification capability is maintained invariably in the research of approximate space (Knowledge Base). Through knowledge reduction, the concept of the classification rule is derived and applied to the unknown data to decide how to make decisions. Without the need to provide the transcendental information beyond the data, Rough Set Theory serves to discover the latent and useful rule, ie.. the knowledge, and its internal connection relations and features in data, In recent years, great success has been obtained concerning knowledge roughness research and the application based on Rough Set and it has become the important branch of soft computing method. It covers pattern recognition, machine learning, decision analysis and policy-making support, knowledge acquisition, knowledge discovery and so on. So, Rough Set Theory has provided the frame for knowledge roughness research.This thesis based on Rough Set Theory elaborates background, the significance and the present situation of knowledge roughness research summarizes the basic thoughts of Rough Set Theory. Through analyzing its origin, nature and the essence of knowledge roughness based on Rough Set Theory, the thesis discusses the knowledge roughness measurement, the structure, the expression and application, showing its certain impetus to the human cognition and problem solution. The thesis has obtained the achievements as follows: (1) based on elaboration of concepts suchas knowledge rough entropy, knowledge granularity, the concept of mutual rough information of knowledge roughness is proposed for the first time;(2) two kinds of expression of knowledge roughness based on the rough set theory are offered, which one rough entropy expression and granularity expression;(3) some relations among the knowledge roughness with the rough entropy and granularity are discussed and established monotonous relations is between the knowledge roughness, between the rough entropy and granularity;(4) based on two kinds of expression, the basic process of attribute reduction in knowledge expression system respectively, the knowledge reduction algorithm is designed respectively, through the example analysis, it indicates that these two algorithms were effective.更多还原