具有非一致性数据预处理的粗糙集特征选择算法

A Rough Set Feature Selection Approach Dealing with Inconsistent Data

大多数特征选择算法面临着对非一致性数据缺乏有效的处理的问题。本文提出了一种处理非一致性数据的方法,采用阈值将非一致性数据做归类处理,当某一类非一致性数据的某个取值比例超过了该阈值,则该类数据都取该值,并只保留一条记录。在此基础上,本文提出了一种改进的基于粗糙集理论的特征选择算法。

Most of feature selection algorithms can not deal with inconsistent data. This article constructs an approach to combine inconsistent data to consistent data. When the ratio of one of the values of a kind of inconsistent data is more than a threhold, the kind of inconsistent data takes this value and only one data is kept. Based on this, the article gives a rough set feature selection approach which can deal with inconsistent data.