【摘要】 粗糙集是粒计算的一种重要方法,数据异构性是大数据的一种特征.针对异构数据问题,探索了粗糙集属性约简的本质,提出了概念属性约简的定义,它兼容值约简、Pawlak约简和并行约简.探究了概念属性约简的性质,提出了异构数据的属性约简方法和概念漂移探测方法.理论分析和示例表明了这些方法的有效性.为粗糙集、粒计算融入大数据的时代潮流提供了一种新方法.更多还原

【Abstract】 Rough set theory is one of important methods of granular computing,and data heterogeneities are one of remarkable characteristics in big data. For data heterogeneities,we define attribute reduction for concepts after investigating intrinsic quality of attribute reducts,which can contain value reducts,Pawlak attribute reducts and parallel reducts. After investigating properties of concept-attribute-reduction,we present a newmethod to reduce redundant attributes and a newmethod to detect concept drift for heterogeneous concepts. Theoretical analysis and examples showthat these methods are valid.This work provides a newtype way for rough set theory and granular computing to integrate into big data.更多还原