The Research Of Attribute Reduction Algorithm In Decision Theoretic Rough Set

Rough set theroy,first proposed by Polish mathematician Pawlak at the end of the last century,mainly for the analysis of uncertainty.The basic and important concepts of rough set theory is attribute reduction, which is a subset of a given set of attributes in the knowledge representation system,it is necessary to maintain a particular property of a single property collection.Attribute reduction is a basic problem in rough set theory,but efficient attribute reduction algorithm is very important and very difficult problem,in reality,has a very important significance.In recent years,researchs have propased many models.such as variable precision rough set models,probabilistic rough set models,which incorporate probabilistic approaches into rough set theory and is built on the traditional Pawlak.These studies greatly improved our understanding of rough set theory and extend the field of application of rough set theory.The results drawn from rough set based classification can be used to make decisions.In traditional Pawlak and probabilistic rough set models,there are two different types of classification rules,positive rules and boundary rules,leading to different decisons and consequences. A positive rule indicates that an object or an object set for sure belongs to or beyond a certain confidence threshold belongs to one decision class;a boundary rule indicates that an objiect or an object set partially belongs to or beyond another weaker confidence threshold belongs to one decision class,which leads to another type of decision.They can be distinguished from the measures such as confidence and risk,these classification rules can be evaluated locally for each individual rule,or globally for a set of rules.This paper introduces the outline of the concept and theoretical knowledge of the rough set theory,analysis of the characteristics ang the present status of the rough set theory;followed by a brief introduction to based on attribute importance.based on discernable matrix,and based on information entropy attribute reduction algorithm ang compare their advantages and disadvantages,and describes the basics of traditional Pawlak and probabilistic these two types of rough set models;The main work of this heesis:(1) On the based of Pawlak rough set models,we introduced the Bayesian decision procedure into decision rough set,so we established the decision-theoretic rough set models,which is based on the loss function.We use the loss function and minimized expection loss to obtain the threshold.lt help our understand the rough set theroy and expend its application fileds.(2) In these models,the positive region is not necessarily monotonous,so we analyzed given certain conditions to limit the positive region so that it is monotonic an under normal circumstances,the attribute reduction algorithm based on attribute importance of the information table,as well as compared with the Pawlak rough ser models attribute reduction results,thus proving the validity of the model.(3) For two different rough set model,a simple assessment of the rules obtained from the different methods of support and loss,so we can get the rules we really want...