Research Of Attribute Reduction Algorithm Based On Positive Region

Rough Set theory is an effective mathematics tool for processing vague and imprecision knowledge, has been applied to many areas successfully including machine learning and data minng and so on..Attribute reduction is one of main topics in rough set theory,is able to reduce the irrelevant attributes from the information system, and form effective rule to help people making the right decision,on the condition of maintaining the ability of classification and decision. A rapid and effective attribute reduction algothrim has not been found at present,so surrounding attribute reduction based on positive region in rough set, the following several works have been done.Analysis found that the existing algorithm for accounting core exist unnecessary sort, so an improved algorithm for accounting core was introduced, experiment results demonstrated that this algorithm is superior of the existing algorithms.Analysis found that the existing algorithm for checking the necessity of each attribute in relative reduction exist shortage, an improved reverse-delete algorithm was proposed,it reduced the number of equivalence classes partitioning.This paper made the following inprovements as for the disadvantages of attribute reduction algorithm based on discernibility object pair set:a new heuristic function was designed and the distributing counting method was provided.The new reduction algorithm did not store discernibility object pair.Because attribute's values may be discontinuous,a algorithm that makes attribute's values continuous distribution based on radix sort was proposed,it can reduce the distributing counting method of time and space complexity.