Research On Bilateral Attribute Values Generalization Reduct In Multi-level Rough Set Model

With the advent of the era in big data,the data with multi-level structure is becoming more common,using the multi-level structure of decision information table(decision table)to realize generalization reduction can reduce the dimension of decision table effectively,while ensure the decision table's classification performance unchanged.And construct a reasonable multi-level structure is the precondition of research generalization reduction.Therefore,it is of great significance to study the multi-level structure construction method and attribute value generalization reduction method of decision table in the era of big data with more and more high-dimensional data.At present,the research on attribute value generalization reduction and multi-level structure construction of decision table is mostly based on the conditional attribute set of decision table,and there are few researches on the multi-level structure of decision attribute,but only considering the attribute value generalization reduction of conditional attribute set can not fully tap the generalization potential of decision table.Every object in the decision table has two kinds of attributes: decision attribute and condition attribute.In reality,many objects' decision attribute also has multi-level structure.Considering only the multi-level structure of condition attribute set,the possible multi-level relationship and generalization potential of decision attribute will be ignored.At the same time,the current research on the multi-level structure construction method is mostly limited to the processing of disordered data,and rarely involves the ordered data with multi-level structure.Based on the above background,this paper conducts the following research:(1)In order to construct the attribute value classification(AVT)which can express the multi-level structure effectively,the VDM-AVT and DAVT learners are proposed and their implementation algorithms are also proposed.The two kinds of learners can automatically construct the AVT of the conditional attribute set and the AVT of the decision attribute according to the decision table respectively,and both of them can process ordered data and unordered data.(2)For the problem of attribute values generalization reduct,this paper first puts forward a multi-level decision attribute the rough set model to study the relationship of the cuts in decision attribute,and then puts forward the multi-level rough set model to study the relationship of the global cuts in decision table.(3)On the basis of the previous model,the definition of decision attribute value generalization and bilateral attribute value generalization reduct is given,and this paper also give the algorithm of decision attribute value generalization and the implementation strategy of bilateral attribute value generalization reduction.(4)The superiority of bilateral attribute value generalization reduct is verified by experiments.