Attribute Reduction Research Of Rough Sets And Fuzzy Rough Sets

Rough set theory is a mathematic tool of dealing with undesigned, half-baked knowledges. It is contrived by Polish mathematician Z.Pawlak in 1982. Now, it is comprehensively used in data mining, intelligent control, pattern recognition and so on. Attribute reduction is one of nuclear contents of Rough set theory. Many scholars apply themselves to researching the algorithm of Rough set attribute reduction.At present, the algorithm of Rough set attribute reduction using attribute significance as heuristic information is the keystone of investigators. So a new attribute significance was given firstly. And then, a new Rough set attribute reduction algorithm was given. In addition, the result of Rough set attribute reduction is not sole. Many attribute reduction results are likely to exist simultaneously. Now, many attribute reduction algorithms define the best reduction algorithm by the least of reduction results. But the correlation of attribute subsets is also important. Based on these views, this paper using attribute relevance weigh the correlation of attribute. And then we choose the least medial relevance result as the final attribute reduction .Many scholars think that we shoud not only consider the classificatory abilities of conditional attribute but also the classificatory abilities of determinative attribute, so a new algorithm was given: Rough set attribute reduction based on system entropy. This algorithm simultaneous considerd the significance of conditional attribute and determinative attribute. It is one of ideal attribute reduction models. This paper modified the definition of system entropy and gave the new algorithm.Dominating Rough set model is an important development of classical Rough set theory. Aiming at the attribute reduction by Dominating relation, the modified algorithm was given. Besides, a new method of attribute reduction using Discernibility matrix was given.Fuzzy rough sets are the important development of classical Rough set. So we should research the attribute reduction of Fuzzy rough sets. This paper put forward a new attribute reduction algorithm based on system entropy, and then, validated this new algorithm.