Research On Data Preprocessing Of Data Mining Methods Which Is Based Upon Rough Set

Due to the classic rough set theory can not deal with the missing data of original data material as well as the data containing the continuous feature, It can be used to gain knowledge to reinforce those data and discretize the continuous feature. So data preprocessing in the rough set theory practice is a very useful process, its result will influence rough set theory practice's efficiency and accuracy. So the research for data preprocessing method in the data mining that is based upon rough set theory has very important meaning.This thesis discusses the question of data reinforce and continuous feature discretization which is based upon data preprocessing of rough set.Firstly, it analyzes and comments the present main discretization calculation, it puts forward a kind of continuous and mixed discretization modified method based on the median sequence division point set to ensure decision table compatibility and to gain reasonable division point.Secondly, it analyzes the features and shortages of present main reinforce calculation. Aiming at the limitation of incomplete systematic reinforce calculation ROUSTIDA, it puts forward a modified data reinforce calculation which can fill up more missing data and tries to avoid decision rule contradiction question which may be caused.