大数据集基于等价类的属性重要性定义和约简

Significance Definition and Reduction of Attributes about Large Datasets Based on Equivalence Classes

属性约简是对大数据集进行数据处理的需要。依据Rough集理论关于属性约简的基本理论和判定条件,以及大数据集启发式约简中属性重要性的重要意义,对大型数据表的启发式属性约简分别提出等价类最大近似度MAAEC和等价类最大差异度MDAEC的属性重要性定义和启发式约简算法。实例演算证明此两法计算有效,在大数据表高维情况下可有效控制计算量,获得属性最小约简集。

It is the need of data processing in attribute reduction about large datasets. According to the reduction basic theory and determinant conditions of Rough Set theory, it brings forward the two definitions of significance of attributes and heuristic reduction algorithms:max|approximation algorithm of equivalence classes (MAAEC) and max|difference algorithm of equivalence classes (MDAEC). They are confirmed by the examples that the two algorithms are efficient, the computing scales are limited and the minimum reducts are obtained.