Research On Data Reduction Method Of Rough Set Theory In E-commerce Application

As a useful mathematics tool, rough set theory can deal with problems of knowledge’s vagueness and uncertainty. The key subjects of rough set theory are computing the core attribute and attributes reduction. The core attribute is the most critical part of all the attributes so that it plays an important role in the whole attributes reduction and even extracting the final rules. The aim of the attributes reduction is to express the information that original information conveys which has been proved NP-hard problem with less and more accurate information by deleting irrelevant and unimportant attributes.When analyzing and comparing advantages and disadvantages of common computing core attributes and attributes reduction algorithms, this paper finds that most algorithms only adapt to consistent decision table, while consider little about the incompatibility of decision table. In order to solve the problems, this paper proposes grading discernibility matrix algorithm that can have different ways to computing core attribute and attributes reduction according to the decision table’s compatibility. When dealing with core attribute in the process of computing core attribute, modified discernibility matrix algorithms from current references are relatively reasonable and effective, so here keep their advantages under the influence of which, grading discernibility matrix algorithm is proposed. The new proposed method defines the final core attribute premised on the possibility that the core attribute is the core computed by the original discernibility matrix and the grading discernibility matrix that is formed by object domains classified by decision attribute values, which is also called domain partition. This paper’s grading discernibility matrix algorithm can compute the core attributes directly when the original method cannot get the discernibility matrix. Both discernibility matrices have their advantages and disadvantages, and to some degrees, they have some connections. Examples show that the proposed algorithm can get the attribute core once the original algorithm cannot do, which demonstrates the grading discernibility matrix algorithm is effective. This proposed algorithm can be used in the research of the attributes reduction, regard the possible core from computing in proper ways as the start point, calculate the reduction set and get the reduction model of the decision table. Examples prove the effectiveness of both algorithms’and this paper does research on the practical application of the two algorithms in the E-commerce data reduction.