D型概率决策形式背景下的规则获取及属性约简

发布时间：2018-09-01 07:57

【摘要】：形式背景是形式概念分析中数据的基本表达形式,它表征了对象与属性之间的二元关系.而决策形式背景是把一般意义下的形式背景的属性集分为条件属性与决策属性两部分,在此基础上研究规则提取与属性约简是决策形式背景的两个重要研究方向.本文针对生活中存在决策属性取值的不确定性这一现象,提出D型概率决策形式背景,定义此背景的协调性,并基于协调的D型概率决策背景分别研究了决策规则获取方法及保持决策规则集相对较优的属性约简问题.主要内容包括:1.基于不确定性决策问题,提出D型概率决策形式背景,定义算子获得概率形式概念,构造相应的概率概念格,并研究D型概率决策形式背景的有关性质;利用包含度的概念,定义D型概率决策形式背景的协调性,并研究《协调D型概率决策背景下的规则获取方法.2.研究保持决策规则集相对较优的属性约简问题,使规则集更加简洁、紧凑;进而研究保持非冗余决策规则集相对较优的属性约简问题.3.研究概率概念格与经典概念格之间的关系;保持格结构不变的约简与保持决策规则集相对较优的约简之间的关系;保持决策规则集相对较优的协调集与保持非冗余决策规则集相对较优的协调集之间的关系.
[Abstract]:Formal background is the basic representation of data in formal conceptual analysis. It represents the binary relationship between objects and attributes. The formal background of decision is to divide the attribute set of formal background into two parts: conditional attribute and decision attribute. On this basis, rule extraction and attribute reduction are two important research directions of formal background of decision. In view of the uncertainty of decision attribute in life, this paper puts forward the formal background of D type probabilistic decision making, and defines the coordination of this background. Based on the background of coordinated D type probabilistic decision making, this paper studies the method of obtaining decision rules and the problem of attribute reduction, which keeps the set of decision rules relatively superior. The main contents include: 1. Based on the uncertainty decision problem, the formal background of D type probabilistic decision making is put forward, the concept of probability form is obtained by defining operator, the corresponding probability concept lattice is constructed, and the properties of formal background of D type probabilistic decision are studied. The consistency of the formal background of D type probabilistic decision making is defined, and the rule acquisition method of < coordinated D type probabilistic decision making background. 2'is studied. In this paper, the attribute reduction problem which keeps the decision rule set relatively superior is studied to make the rule set more concise and compact, and the attribute reduction problem of preserving the non-redundant decision rule set is studied. The relationship between probabilistic concept lattices and classical concept lattices is studied, and the relation between the reduction with invariant lattice structure and the relatively superior reduction of preserving the decision rule set is discussed. The relationship between the relatively optimal coordination set of preserving the decision rule set and the relatively optimal coordination set of preserving the non-redundant decision rule set.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;O225

【相似文献】