| The rapid development of network information technology and computer technology makes data expand rapidly,and it becomes especially important to find valuable information in the huge amount of data.Rough set theory provides an effective method for solving information problems such as incompleteness and uncertainty.It does not require any additional information or prior knowledge and can be used to approximate uncertain or ambiguous concepts by defining upper and lower approximations about them.Classical rough set theory studies incomplete information systems and uses equivalence relations to divide the domain of the argument into unrelated equivalence classes.However,the practical situation is that the application of classical rough set models is limited due to data measurement errors,dynamic changes in data values,omission or increase of objects,missing attributes,etc.This paper is a study of incomplete information systems and a rough set model based on non-symmetric restricted similarity relation as the research objective,the following results were obtained.(1)A rough set extension model based on non-symmetric restricted similarity relation is proposed.The extension conditions of the rough set extension model are improved by redefining the similarity classes and thus obtaining new upper and lower approximations.It is shown that the non-symmetric restricted similarity relation R is related to the tolerance relation T,the restricted tolerance relation L and the similarity relation S in a certain way.A comparison of the performance between these rough set extension models is made through an example that highlights the advantages of this extension model.(2)Under the incomplete information system,the discernible matrix with clear and intuitive characteristics is selected,and the constraints of matrix elements are improved by using the proposed non-symmetric restricted similarity relation to obtain a new reduction algorithm.Combined with a fire emergency example,the reduction attributes can be obtained by simply constructing the identifiable matrix and then deriving the identifiable function,which improves the efficiency of the simplification.(3)A new decision rule formulation algorithm is proposed,which combined the rough set expansion model of non-symmetric restricted similarity relation with the binary discriminable matrix and then simplifies the binary discriminable matrix to obtain a new matrix.The binary discriminable matrix is optimized while the small binary discriminable matrix corresponding to each decision rule is attribute reduced.The simplified decision rules are derived and combined to obtain the final set of decision rules.The advantage of this algorithm is that it is only necessary to find all pairs of instances containing a certain instance in the improved binary recognition matrix,construct its corresponding binary recognition matrix,and the rules can be extracted quickly. |
