Study On Attribute Reduction In Incomplete Information Systems Based Rough Set Theory

Presently, one of the frontest directions in information decision-making research is the data mining in which many kinds of theories and technologies are combined such as the machine learning, the neural network, the pattern recognition, rough set,and so on.Rough set theory is a kind of mathematical method used to process imprecise and indefinite knowledge.In recent years,rough set theory is noticed highly because of its widely applications in such the fields as pattern recognition,data mining,and decision analysis.In the classic rough set theory, equivalent relationship is taken as foundation and all the attribute value of every object in the domain is supposed to be known.However,incomplete information system exists massively in data mining, which restrictes the application of classic rough set theory since the traditional equivalent relationship no longer exists. Therefore, how to carries on data mining from an incomplete information system, especially carries on data mining while not changing the incomplete information system mentioned above is a significant research task.Attribute reduction methods of incomplete information systems are the main contents to be researched in this paper.Firstly,data mining and classic rough set theory are elaborated systematically.Secondly,some extended models of rough set theory and the attribute reduction algorithms of incomplete information system are analysed and compared.Furthermore,a reduction algorithm of incomplete information systems is presented based on both granular computing theory and rough set theory.Finally,a new measurement method for uncertainty about knowledge and rough set of incomplete information systems is presented. The influence of boundary region to uncertainty is considered by the new definition which provides an effective tool for the attribute reduction and the knowledge acquisition of incomplete information systems.