属性约简中的近似问题研究

Study on Approximation Problem of Attributes Reduction

现有相对约简实际上只是在保持所关心范畴的下近似不变的前提下对属性进行约简,这对于不协调信息系统来说,可能会使所关心范畴的上近似变大,从而使其近似精度减小,导致有用信息的丢失.为此,针对属性约简中的近似问题进行了研究,引入近似相对约简概念,利用近似相对约简替代现有相对约简来处理不协调信息系统,以克服现有相对约简在不协调信息系统中所带来的缺陷.同时,证明了在协调信息系统下,条件属性集相对决策属性集的近似相对约简和相对约简是等价的,并提出了一种基于近似相对约简的简便属性约简算法.

The prerequisite of available relative reduction is to keep lower approximation of category unchanged, but then it may increase upper approximation of some category which we are interested in, and decrease approximation precision for inconsistent target information system. Some useful information may be lost. The main objective of this paper is to introduce a new concept of attributes reduction based on the relationship of attributes reduction and approximation sets. It is referred to as approximation relative reduction, which replaces available relative reduction and overcomes the drawback of the available relative reduction in inconsistent system. At same time, the effect of approximation relative reduction is same as that of available relative reduction in consistent target information system, as is proved. A kind of attributes reduction approach based on approximation relative reduction theory is presented in this paper.