Research On Rough Set Theory And Its Application In Data Mining

Data mining aims to find out the way of abstracting automatically and intelligently valuable information or knowledge from a huge amount of data. At present this is an active research field. In recent years, in contrast to the rapid development of our national information construction, the technology of the automatic acquisition of knowledge has become its bottleneck. Therefore, works in this area are theoretically important and practically valuable.This thesis deals mainly with the methods and technologies of data mining based on Rough Set Theory.With respect to data discretization, the focus is firstly placed on the discussion and analysis of advantages and disadvantages different methods bear in dealing with this problem, and then a new method is put forth for the solution of data discretization aiming at continuous attributes in decision table. By using the classification information provided by decision attribute, this method not only avoids the complex clustering operation but also can output a result with high data consistency.Then the thesis discusses the extension of Rough Set Theory in order to deal with incomplete information system. In this chapter, we bring forward a Variable Precision Rough Set Model. Because of the introduction of fuzzy technology, it can go on with knowledge reduction effectively on different conceptual level.In rule extraction, we give an algorithm of mining Weighted-Decision-Rules In Decision Table, which is based on interests of users. It uses the concept of support proposed in this paper to select rules most interested from the candidate. It suggests a new way of data mining.Data pretreatment, knowledge reduction and rule abstraction are three important problems in the research of Rough Set Theory. The work of this paper on them is valuable. But data mining is in a booming stage and there are many problems worth studying on the application of Rough Set Theory in this field. Our work is just a beginning, and related work needs to be further developed.