An Expand Research On Attribute Reduction Theory And Methods For The Fuzzy Information System

Attribute reduction is an important part of rough set theory, and also a hot issue of knowledge discovery of information system. With rapid development of the information age and the constantly derivatives of massive data, there is more and more attention on data mining and information processing of complex systems. This paper follows with interest in fuzzy information systems. Some existed problems of fuzzy information systems such as the knowledge representation, the similar description of objects and the relevance characterization of attributes, etc., will be discussed by means of the rough set methods connecting with fuzzy set theory and granularity analysis. A generalized research of attribute reduction theory and methods for the fuzzy information system will be conducted.This paper is divided into five chapters. Chapter I briefly illustrates the research background, the research status, the research significance and the basic notions involved in the paper. Chapter II first proposes the notion of condition attribute similarity and decision attribute similarity in terms of the fuzzy binary similarity relation, and a similar class matrix of attributes is established. On this basis, the correlation between condition attributes and decision attributes is considered, by which a relative comparison matrix about conditions similarity and decision similarity is constructed. Therefrom, the definitions of attribute coordinated set and attribute reduction set are introduced. Furthermore, associating knowledge granularity with attribute discemibility and attribute correlation, the importance of condition attributes relative to decision attributes is analyzed. As a result, this study obtains a way of fuzzy attribute reduction by utilizing the similarity comparison. Chapter Ⅲ, according to fuzzy similarity matrix produced by similarity relation, the fuzzy similar class is discussed. Furthermore, the upper and the lower approximation operators for an object subset are defined respectively. The fuzzy positive domain class of a condition attribute set to a decision attribute set is thus induced, which leads to the study for importance measure of attribute set. Finally, an algorithm of fuzzy attribute reduction is put forward based on the fuzzy positive domain class. Chapter IV, the concept of similarity cut matrix is first introduced by the condition similarity and the decision similarity. A new definition on attribute coordinated set and attribute reduction set is induced in terms of the similarity cut matrix. Moreover, this part discusses the relevant attribute reduction theory by associating with the elements sum of similarity cut matrix. In the last, a new attribute reduction method is introduced under the similarity cut matrix. Chapter Ⅴ is a summary of this paper.