The Research Of Attribute Reduction Method Based On Superior Relationship

Nowadays,with the rapid development of information technology,the application of big data has become more and more extensive.Various data have penetrated into all walks of life in society.As a result,massive data has become the most valuable asset in today's society.Many companies started to build their own database in order to create new value through data processing.However,the continuous increase in the amount of information has directly led to the rising cost of information processing.Therefore,how to accurately analyze and process high-dimensional data has become a new research hot spot.Attribute reduction,as a method to deal with the uncertainty of data,has aroused widespread concern of scholars.It can efficiently process high-dimensional data by deleting the redundant features in the research data retaining the necessary features,and keeping the classification ability of the system unchanged.Using attribute reduction before using data can effectively simplify the process of high-dimensional data and obtain valuable information from the data.In this paper,several new superior relationship models are given,and the characteristics of the model are further studied.For these several models,the concepts of the lower approximation and positive domain of the decision attributes are defined respectively.The dependence degree function is further constructed to reflect the the classification ability of the.In the process of experiment,it is found that the common superiority relationship has the drawbacks that are inconsistent with the actual situation when comparing the superiority of samples.This is because in the process of calculating intersections with multiple attributes,two samples are needed to have the same superiority in all attributes,the results can be obtained.For this reason,the method of transposition summation is used in this paper instead of the traditional method of calculating intersections,and the concepts of the lower approximation of decision attributes,the dependence degree function and the attribute importance are redefined.The effectiveness of the proposed algorithm for attribute reduction is verified by numerical experiments.