一种带局部回溯的属性约简算法

1 引言粗糙集理论自1982年由波兰科学家Z.Pawlak提出发展到现在,知识约简已经成为其研究的核心内容之一。知识约简包括属性约简和属性值约简,属性约简是对整个知识库而言的,在保证信息量不减少的前提下,去掉冗余的属性;而属性值的约简是针对每条信息(决策)的描述而言的,在不影响对其分类或决策的能力的情况下,不考虑某些属性的取值。本文所提到的约简如果没有特别说明,特指属性的约简。一般来讲,约简是不唯一的,人们当然希望能够找到具有最少属性的约简,即最小约简。遗憾的是,求解最小约简是NP完全问题,任何非穷举的算法都不能保证所得结果是最优的,这时不得

The Attribute Reduction Algorithm with Local Retrospect

In this paper, the existing attribute reduction algorithms in rough sets are classified as members of TD-al-gorithms (Top Down); then a new reduction algorithm of BU(Bottom Up) is proposed; after the comparison to theideas of the two groups of reduction algorithms, an effective attribute reduction is created through adding proper localretrospect to the traditional methods; at last an experiment is given to show the validity of this new technique.