Study On Covering Rough Sets And The Algorithm Of Attribute Reduction

With the advant of globalization of information,companies has accumulated vast amounts of data in the process of production and management for the purpose of improving their competitive advantage. How to get valuable information from these data is one of the key problems to be solved in the field of data mining. Rough set theory bas been proposed by Pawlak as a tool of data analysis for dealing with uncertainty and incomplete knowledge in data mining. The core concepts of the rough sets are lower and upper approximations based on eqivalence relations, and the knowledge which hidden in the information system can be expressed in the form of decision rules by means of these two concepts. Since classical rough set based on equivalence relations has been limited in complex information system,it has been generalized by many researchers. Z. Bonikowski establish a covering rough set theory by using the covering of the domain to construct the lower and upper approximations. EricC.C. Tsang defined induced covering rough set theory,which futher enriched the covering rough set theory. The duality does not hold in the induced covering rough set theory,so in this paper the lower and upper approximation operator are be modified so that they are dual. This paper also discusses the related properties and relationships of this two models.The classical rough set is not only susceptible to the impact of the data containing noises and also having poor tolerance of fault. Ziarko variable precision rough set model is introduced a parameter to deal with these problems, but this theory is still limited to eqivalence relations of the domain. This paper define induced covering variable precision rough set model make reference to the idea of variable precision rough set model,and discuss its related concepts and algebraic properties.Knowledge Reduction (attribute reduction) is a key issue in the field of data mining, and rough set theory as a data mining tool has been committed to the attribute reduction algorithm. Most studies focused on the approximation operator and properties of covering rough set theory in the past.William Zhu proposed a method of attribute reduction basing on Z. Bonikowski covering rough set model. The method is only to eliminate redundant elements of a cover as a technology to eliminate redundant data in a knowledge base, so its use has been extremely limited and this article propose a new method of attribute reduction combining the theory of evidence with the induced covering rough set theoryⅡ,finally this paper apply the algorithm to Multi-value information system.