Research On Structure Features Of Covering Information System

With the continuous development of information technology and database technology,massive information is stored in different form. In order to effectively use the importantinformation behind these data, data mining is gradually developing. Data attributes arenot equally important, even some of the attributes are redundant. These redundant data notonly occupy a large amount of storage space, more bring many inconvenience toknowledge discovery and decision making. Therefore, refining the database attributes(attribute reduction) become an important part of data mining technique.Due to it is often difficult to get accurate classification in practical problems,covering information system has been used widely in data mining. Its main idea is to keepthe classification or decision-making ability of knowledge base without changing and deletethose uncorrelated or unimportant attributes. But these methods all have higher time andspace complexity and cannot reduct large database effectively. Therefore, the research ofstructure characteristics about covering information system is particularly important.To the above problems, firstly, this paper proposes the concept of fuzzy covering bybuilding a fuzzy similarity relation. By analyzing the impact of precision change oncoverings ascertained by fuzzy sets, we arrive at some important conclusions; Secondly,we put forward the determination method of δ neighborhood covering. And again,summarizes several common method of cover reduction, and analysis their characteristicsand shortcomings. We propose the idea of minimization covering by using the propertiesof sets; Furthermore, in order to reduce the redundancy of covering, we come up with theidea of network topology. And we build a attribute reduction method of coveringinformation system based on the network topology. At last, combined with concreteexample, we prove the characteristics and advantages of these methods. And we illustratethat the two methods of determining covering given by us have a certain practicalrationality and validity. They can deal with attribute reduction problems of databaseeffectively.