基于邻域粗糙集组合度量的混合数据属性约简算法

属性约简是一种重要的数据挖掘方法。为了对混合型信息系统达到更好的属性约简性能,提出一种邻域组合度量的启发式属性约简算法。邻域依赖度是构造混合信息系统属性约简的常用方法,根据粒计算的视角,在混合信息系统中提出邻域知识粒度用于评估属性的粒化能力。将邻域依赖度与邻域知识粒度进行结合,提出混合信息系统下的邻域组合度量,并将该度量方法作为启发式函数,提出一种属性约简算法。实验分析表明,该算法比混合信息系统的其他相关属性约简算法具有更高的约简性能。

ATTRIBUTE REDUCTION OF MIXED DATA BASED ON NEIGHBORHOOD ROUGH SET COMBINATION METRICS

Attribute reduction is an important data mining method.In order to achieve better attribute reduction performance for mixed information system,this paper proposes a heuristic attribute reduction algorithm based on neighborhood combination metric.Neighborhood dependency was a common method for constructing attribute reduction in mixed information system.According to the perspective of granular computing,the neighborhood knowledge granularity was used in the mixed information system to evaluate the granulation ability of attributes.By combining neighborhood dependency with neighborhood knowledge granularity,we proposed a neighborhood combination metrics in hybrid information system,and this method was taken as a heuristic function to propose a new attribute reduction algorithm.Experimental analysis shows that the proposed attribute reduction algorithm has higher reduction performance than other related attribute reduction algorithms in the mixed information system.