Studies Of Some Methods Based On Reduction Attribute Of Rough Set

With the continuous progress of society and the development of science and technology, the age of information provides people with a variety of convenient, but the large data sets also makes people feel confused. It will be a very important research topic to find useful knowledge from the complicated data. If you want to get useful information effectively, you have to accelerate the pace to discover knowledge. Rough set theory doesn’t require a lot of prior knowledge, it has made remarkable achievements in artificial intelligence and other fields. The main idea of rough set theory is based on the existing knowledge and knowledge base to describe the uncertainty problem. In general, we use the upper and lower approximation set to describe fuzzy knowledge, so that we can have deep understanding on fuzzy knowledge.This paper mainly studies the algorithm based on rough set. Concrete research content is as follows:First, this paper introduces the basic knowledge of rough set theory. It uses the basic definition of knowledge and knowledge base as starting point, then it gives the concept of information decision and the definition of upper and lower approximation set and the relevant nature.Second, the attribute reduction algorithm is studied. This part is the focus of this article core content. According to the definition of attribute importance and gives the relevant algorithms. The article first lists an example and the result is feasible,but when the article lists the second example and the result is infeasible. According to analysis of the decision table. The result is that the first table is compatible,but the second is incompatible. So we can confirm that this algorithm is suitable for the compatible system not for the incompatible system. Then the article comes up with the improved algorithm which removes the incompatible objects when calculating the dependence,and we can get accurate reduction set. Then the article also analysis the advantages of the algorithm, according to compare with others. The first advantage is that it can solve the problems which are compatible or incompatible; the second is that it can solve the attribute more conveniently. Only we do is that we should observe whether the importance of the attribute is greater than zero.Finally, the article studies attribute reduction algorithm under the fully discreteinformation system. Then it puts forward the reduction algorithm based on mutual information gain rate and its shortage. Then, we proposes a improved algorithm. The improved algorithm not only considers its own information entropy but also considers the change of mutual information when increasing attributes in reduction set. According to the improved algorithm,we can obtain the reduction set which is better.