Rough set theory is one of the theoretical basis of algorithm on data mining. And it has been generalized ever since be proposed. The main work of this paper is to promote the Pawlak rough set model.First of all, in order to solve the problems the classical model face in dealing with information systems which contain continuous-valued attributes or inconsistence data, Pawlak Rough set theory was prompted in this paper. Fuzzy Rough set Based Fuzzy Coving was constructed as a new model. And its properties were explored. It was discovered and proved that the new model is more general than Pawlak Rough Sets, Rough Fuzzy Sets, Covering Rough Sets. Then, we do some axiomatic argument about it and found some important natures of the model, on account of which we give the axiomatic definition of the upper and lower approximation operators. Next, combined with the standard data which is widely used in data mining research--the quality of the wine data, we perform an experiment on attributes reduction. The experiment shows that the reduction effect.Secondly, we try to find a solution from the angles of domin decomposing, to reducing the complexity of Rough Set-computing in the fuzzy knowledge space of high-dimension. In this section, we first researched the decomposing of double universe. And then, constructed a model of product fuzzy approximation space. The potray and decomposition of product fuzzy rough set is discussed, and a characterization of upper (lower) approximation of crisp decomposable sets is given, from the angles of λ-cut approximation spaces, as well as the algorithm on decomposing the fuzzy decomposable sets. |