两种粗糙集模型下的属性约简方法研究

发布时间：2018-05-28 15:15

本文选题：粗糙集 + 属性约简　；参考：《江苏科技大学》2017年硕士论文

【摘要】：粗糙集理论作为智能信息处理的一种重要方法,由波兰科学家Pawlak首先提出,引起大量学者关注并对其开展研究。在研究过程中,学者们为了打破粗糙集应用场景的局限性,提出了一系列的粗糙集扩展模型。值得一提的是,其中有两个粗糙集模型应用十分广泛,即:模糊粗糙集模型、决策粗糙集模型。前者用于改善粗糙集理论在处理模糊问题的乏力状况,使得粗糙集理论在模糊性问题的处理方面也有了较好的能力;后者是一个借助贝叶斯风险决策理论进行改进的粗糙集模型,它给粗糙集引入语义解释,极大地缓解了粗糙集解决问题缺乏科学语义支撑的尴尬局面。同时,它也消除了经典粗糙集零错误容忍率和现实应用中存在错误率的矛盾。随着这两个改进粗糙集模型的广受欢迎,作为粗糙集理论的核心内容之一的属性约简,其研究的价值也变得越来越大。针对以上研究问题,本文拟从模型的理论和应用两个方面开展该研究工作。主要研究内容如下:(1)针对模糊粗糙集模型,我们从测试代价出发,根据模糊粗糙集模型的特性,分析测试代价敏感的属性约简方法的实现方法,并给出两种不同的算法思想,定义出其相对应的基于降低测试代价原则的算法,并通过实验对理论进行验证这两种算法在处理该问题的效率。从实验结果我们可以发现,遗传算法在处理该问题的时候在不考虑时间的情况下能得到更好的结果。(2)针对决策粗糙集模型,我们从决策规则的角度出发,对决策保持属性约简和决策单调属性约简方法进行分析。将局部约简的方法引入决策粗糙集模型下,定义出局部决策保持和决策单调属性约简的方法,分析方法的可行性。然后再给出算法思想,并依据算法思想,进行实验,进一步对理论的实际可行性进行验证。从实验结果我们可以发现,局部的决策保持以及决策单调约简算法分别在降低冗余属性和获取决策规则两个方面相较于全局属性约简具有更好的处理能力。
[Abstract]:As an important method of intelligent information processing, rough set theory was first proposed by Polish scientist Pawlak. In order to break the limitation of rough set application scenario, scholars put forward a series of rough set extension models. It is worth mentioning that two rough set models are widely used, namely, fuzzy rough set model and decision rough set model. The former is used to improve the weakness of rough set theory in dealing with fuzzy problems, which makes rough set theory have better ability in dealing with fuzzy problems. The latter is an improved rough set model based on Bayesian risk decision theory, which introduces semantic interpretation to rough set, which greatly alleviates the awkward situation of rough set problem solving lacking scientific semantic support. At the same time, it eliminates the contradiction between the zero error tolerance rate of classical rough set and the error rate in practical application. With the popularity of these two improved rough set models, attribute reduction, which is one of the core contents of rough set theory, has become more and more valuable. In view of the above problems, this paper intends to carry out the research from the theory and application of the model. The main research contents are as follows: (1) in view of fuzzy rough set model, we analyze the implementation method of attribute reduction method which is sensitive to test cost according to the characteristics of fuzzy rough set model, and give two different algorithms. The corresponding algorithms based on the principle of reducing test cost are defined, and the efficiency of the two algorithms in dealing with this problem is verified by experiments. From the experimental results, we can find that the genetic algorithm can get better results when dealing with this problem without considering the time. (2) for the decision rough set model, we start from the angle of decision rules. The methods of decision preserving attribute reduction and decision monotone attribute reduction are analyzed. The method of local reduction is introduced into the decision rough set model, and the methods of local decision retention and monotone attribute reduction are defined, and the feasibility of the method is analyzed. Then the idea of the algorithm is given and the experiment is carried out according to the idea of the algorithm to verify the practical feasibility of the theory. From the experimental results we can find that local decision retention and decision monotone reduction algorithms have better processing ability than global attribute reduction algorithms in reducing redundant attributes and obtaining decision rules respectively.
【学位授予单位】：江苏科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】