中文摘要：

从相对核的角度,提出了一种新的属性约简方法.首先,求出条件属性相对决策属性的相对正域,然后根据相对正域求得属性的相对核.用这些相对核属性对论域进行划分,在对论域划分后,将可以完全正确的分类删除,减小论域,如此迭代下去,直到论域完全划分,最后求出这些核属性并集,去除并集的冗余信息,即可得到属性约简集.该方法可直接利用核属性来对论域进行划分,不用再计算每个属性的重要度,减少了计算量,在每次迭代的过程中,减小论域,缩减搜索空间,降低了时间复杂度.

英文摘要：

From the point of view of relative core,this paper proposes a new attribute reduction method.Firstly,the condition attributes and the decision attributes are used to calculate the positive domain.Then,the relative core of the condition attributes is got based on the positive domain.Secondly,the samples are divided with these relative core attributes.At the end of this division,the samples that can be divided correctly is deleted.And then the samples are reduced.This iteration continues until the samples are completely divided.At last,the union of relative core is got and redundant information is removed,and then attribute reduction set is obtained.This method can use core attributes to divide the samples directly.No longer to calculate the important degree of each attribute,and then the amount of computation are reduced.In each iteration process,the samples,the search space and the time complexity are reduced.