【作者】 纪怀猛；

【导师】 罗可；

【作者基本信息】 长沙理工大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 粗糙集理论是20世纪80年代初由波兰数学家首先提出的一种刻画不确定和不完备知识的数学工具,能有效地分析和处理不精确、不一致、不完整等各种信息,并从中发现隐含的知识,揭示潜在的规律。经典粗糙集理论以完备系统为研究对象,以等价关系为基础,通过等价关系将论域划分为互不相交的等价类;然而,在现实生活中,由于数据测量的误差,对数据理解或获取的限制等原因,使得在知识获取时往往面临的是不完备系统,即可能存在部分对象的一些属性值未知的情况,这就极大地限制了粗糙集理论向实用化方向发展。因此,如何从不完备决策系统中应用粗糙集理论获取知识己经成为当前的一个重要研究课题。本文首先综述了数据挖掘的原理和现状,从数据挖掘和知识分类的角度出发,探讨了数据挖掘的相关概念、工作步骤和关键技术。深入分析了粗糙集的基本理论以及粗糙集理论在不完备系统中的拓展。然后基于相容关系研究了把信息论和集合论应用于不完备决策系统属性约简的方法,以及粗糙集理论与遗传算法相结合的不完备决策系统属性约简算法,通过实验数据验证所提出的约简算法,得到不完备决策系统相对应的约简结果。接下来研究了在属性约简后的决策表中提取最优规则的算法并通过实例分析验证了算法的正确性。本文最后设计和开发一个基于粗糙集理论的不完备决策系统的数据挖掘模型,应用本文算法进行属性约简和最优规则提取,部分实现该模型的功能。更多还原

【Abstract】 Rough set theory, introduced by Pawlak Z. in the early 1980s, is a new mathematical tool used for dealing with vagueness and uncertaint information to discover implicit knowledge or reveal latent laws. Classic rough set theory based on equivalence relation takes complete system as object of study , and divides the region into some non-intersect equivalence class; But, in the real life, because of the errors in data measuring, understanding of data, or the restriction in data collection , it can make the decision-making system incomplete, that is, value of attribution of some objects is unknown, which restrains development of the theory to practical direction. So how to acquire knowledge from incomplete decision-making system has been a crucial research topic recently.The paper, firstly, sums up the principles and reality of data mining, and discusses the corresponding concepts, working steps and key technologies about data mining from the viewpoint of data mining and knowledge classification, making a deep analysis about basic theories and extension in the incomplete system. Then, based on compatible relation, study of combining information theory with set theory and study of combining rough set theory with genetic algorithm for reduction have been made. Algorithms for reduction are validated by experiments. It shows that these algorithms can find corresponding reduction results. Next it puts forward an algorithm that can acquire directly decision-making rules in decision-making table. All algorithms are validated by experiments, the results of which show that these algorithms turn out right.At last, it establishes a system for data mining in incomplete decision-making system, applying the algorithms put forward in the paper, for attribution reduction and optimum rule extraction to fulfill the functions of the model.更多还原