摘要： 针对以依赖性作为属性重要性度量的约简算法效率较低、不能有效处理不一致信息系统的问题，提出一种时间复杂度为O(|A|2|U|)、基于错误分类率(ER)的快速约简算法。根据等价类计算的包含关系和正区域与属性个数的单调关系，采用ER作为属性重要性的度量。在UCI数据集合上测试该算法，结果证明了其有效性。

关键词: 粗糙集, 依赖度, 错误分类率, 快速算法

Abstract: Aiming at the inefficiency problem of existing reduction algorithms which use dependency to evaluate the quality of an attribute, and can not deal with inconsistency system, this paper proposes an algorithm based on Error Rate(ER) whose complexity is O(|A|2|U|) to compute the equivalent classes. This algorithm uses the include relationship of equivalent classes and the property that positive region increases with the amount of attributes. It introduces ER as the measure of an attribute. Experimental results with UCI data show that this algorithm is efficient.

Key words: rough set, dependency, error classification rate, fast algorithm

中图分类号: 

• TP301.6