基于粒划分粒重叠发现原理的特征选择方法研究

发布时间：2018-04-05 15:02

本文选题：粒计算　切入点：形式背景　出处：《燕山大学》2012年硕士论文

【摘要】：粒计算是一种基于问题概念空间划分的新的智能计算理论和方法；粒计算能够有效分析和处理模糊、不精确、不一致和部分真值的问题；粒计算是覆盖了所有有关粒度的理论、方法论、技术和工具的研究；粒计算不仅是模糊集理论、粗糙集理论、商空间理论等的超集，而且已经成为人工智能领域中一个重要的研究分支，具有广泛的的应用前景。首先，本文基于粒计算理论和形式概念分析理论，讨论了粒计算理论中有关粒、粒度、粒层等基本概念，，以及形式概念分析理论中有关偏序、形式背景等的基本概念。基于中医证名、证素所形成的形势背景，进行了“粒”和“粒空间”的构造，以此来作为概念格的化简。最终，在“粒”和“粒空间”构造的基础之上形成一种满足人们需求的、背景复杂的、层次清晰的、化简简单的基于粒计算的属性约简方法，此方法所形成的粒具有层次关系。比如：对于属性投影空间，最细的是由M决定的粒空间，这时每一个概念就是一个粒；最粗的是由(?)决定的粒空间，这时所有概念是一个粒。其次，论文提出了一种新的形势背景优化的方法、属性偏序结构图构造的方法，形成了无交叉、层次清晰明了的粒层粗细的概念格粒度划分图。最后，论文依据有关肾病证类的实际事例，进行了属性偏序结构图的构造。换句话说，就是进行了粒度层次的划分或者等价关系的构造，以便于进行肾病证类的确诊或者规则的提取。
[Abstract]:Particle computing is a new intelligent computing theory and method based on the partition of problem concept space. Particle computing can effectively analyze and deal with fuzzy, inexact, inconsistent and partial truth problems.The research of methodology, technology and tools, granular computing is not only the superset of fuzzy set theory, rough set theory and quotient space theory, but also has become an important research branch in artificial intelligence field, and has a wide application prospect.Firstly, based on the theory of granular computing and formal conceptual analysis, this paper discusses the basic concepts of grain, particle size and grain layer in granular computing theory, as well as the basic concepts of partial ordering and formal background in formal conceptual analysis theory.Based on the situation background of TCM syndromes and syndromes, this paper constructs "grain" and "grain space" to simplify the concept lattice.Finally, on the basis of the construction of "grain" and "grain space", a kind of attribute reduction method based on granular computing is formed, which meets the needs of people, has a complex background, a clear hierarchy and simple simplification.For example, for attribute projection spaces, the finest is the grain space determined by M, where each concept is a grain, and the coarstiest is a grain.)Determines the grain space, at which point all the concepts are a grain.Secondly, this paper proposes a new method to optimize the background of the situation, the method of constructing the attribute partial order structure diagram, which forms a clear grain-level concept lattice granularity partition graph without cross and clear layers.Finally, according to the actual cases of nephrotic syndrome, the structure of attribute partial order structure is constructed.In other words, granularity level partition or equivalent relation is constructed to facilitate diagnosis or rule extraction of nephrotic syndrome.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0;TP18

【参考文献】