Research On Data Analysis Model And Algorithm Based On Fuzzy Rough Set

Today we are in an era of informationization.The era of informationization is a general trend of development in the present age.With the development of network information technology,we are at an unprecedented pace of development.With the reality of daily life and big data problems of our life waiting to be solved.For feature selection,fuzzy rough set is an important classical rough set model.Fuzzy rough sets use fuzzy dependent function as feature selection standard.However,this standard function can not guarantee the minimum classification error while maintaining the maximum membership of a sample to a decision class.In this paper,a new feature selection criteria is introduced to overcome this weakness.In order to characterize the classification error ratio,first introduce the class of fuzzy binary relations to construct the fuzzy upper approximation and lower approximation.Then,introduce the new dependency concept: use misclassification rate and inner product dependence to describe classification errors.Based on this,we propose new feature selection criteria for measuring importance of candidate attributes.The proposed criterion can guarantee the minimum classification error while maintaining the maximum dependence function.The algorithm and experimental results show that the feature selection algorithm in this paper outperforms some other classical algorithms,especially for data sets that show a large overlap between different categories,which not only reduces the complexity of feature selection but also improves the classification accuracy,have a certain practical significance.