Research On Incomplete Information Filling And Attribute Reduction Based On VPRS Model

The rough set theory model is a powerful tool for data mining which is widely used in machine learning,process control,knowledge acquisition,pattern recognition,etc.Rough set theory can find hidden information in the data table through attribute reduction.While in the actual production process and life,the data collection process often encounters data missing,uncertain,and incomplete,so that the incomplete data table finally collected cannot directly apply classical rough set theory to attribute reduction.Therefore,this paper starts with the filling and attribute reduction of incomplete decision information systems and conducts related research on it.This paper takes incomplete decision-making information system as the object,studies the filling method of its missing attribute value,and proposes the definition of variable precision rough set attribute reduction based on maximum positive region,and applies it to the filled information system to carry out attribute reduction.The details are as follows:1.Briefly described the domestic and foreign research progress on filling methods of incomplete information systems,classified and explained the attribute reduction of existing classical rough sets and variable precision rough sets.In order to design a new filling method for incomplete information systems,and apply variable precision rough set attribute reducti-on to the filled information system,so as to determine the main research content of this dissertation.2.Improved the discriminant matrix of incomplete decision-making information system,and proposed the concept of contribution matrix,given the definition of attribute contribution rate,attribute completion rate and so on based on it.Then,the relative contribution rate and relative completion rate of the attribute are combined with the parameter ?,? to define the attribute importance.3.The definition of important attributes and unnecessary objects is given on the basis of the importance of attributes based on parameters ? and ?,and the corresponding missing value filling method is given according to the frequency of attribute values in the decision table.This method can transform incomplete decision information system into weighted decision information system.4.Aiming at the limitation of the existing attribute reduction definition,the concept of maximum positive region in variable precision rough set model is proposed,and the corresponding attribute reduction definition is given.At the same time,the relationship between attribute reduction based on maximal positive region and existing attribute reduction is discussed.For the convenience of application,the general algorithm of attribute reduction based on maximum positive region and the attribute reduction algorithm of maximum positive region VPRS model based on genetic algorithm are given respectively,and their advantages and disadvantages are analyzed.5.For the weighted decision information system,the attribute reduction definition of the VPRS model that keeps the relative positive region unchanged is given first,and then the attribute reduction definition and algorithm of the VPRS model based on the maximum positive region wereproposed.Finally,two databases in UCI were selected to verify the feasibility and effectiveness of the proposed algorithm.