摘要
基于差别矩阵属性约简方法获得的约简并不能保证与正区域约简一致,并且在处理高维大数据集时将消耗过多的存储空间.为此,首先对决策表进行简化,引入属性布尔差别矩阵及其核属性和属性约简定义,同时证明了该约简与正区域约简是一致的.然后,基于属性布尔差别矩阵,设计正向启发式属性约简算法;为了进一步减少算法的空间开销,引入Swapping技术,并给出反向启发式属性约简算法.最后,实例和实验结果表明所提出的约简算法是正确的、高效的.
When some attribute reduction methods based on discerniblity matrix deal with inconsistent and high-dimensional data sets,the shortcomings of the attribute reduction methods will appear,which are that the reduction is possible to be inconsistent to the positive region reduction and the space-consuming is more larger. Hence,the method of simplified decision table is proposed firstly. The definition of attribute boolean discernibility matrix is provided,and the corresponding definitions of core attribute and attribute reduction are put forward. It is proved that attribute reduction acquired from the definitions is equivalent to the attribute reduction based on positive region. Then,the forward heuristic attribute reduction algorithm is designed. In order to reduce space cost,the swapping technology is introduced into the attribute reduction and an improved algorithm is present,which is backward heuristic attribute reduction algorithm. Finally,both of the example analysis and experiment results show that attribute reduction algorithms proposed are correct and efficient.
出处
《小型微型计算机系统》
CSCD
北大核心
2014年第7期1620-1624,共5页
Journal of Chinese Computer Systems
基金
安徽省自然科学基金项目(1308085QF114)资助
安徽高校省级自然科学研究项目(KJ2012A212
KJ2013A015)资助
滁州学院优秀青年人才基金重点项目(2013RC003)资助
滁州学院科学研究项目(2011kj003Z)资助
关键词
粗糙集
属性布尔差别矩阵
属性约简
核属性
Swapping技术
rough set
attribute boolean discernibility matrix
attribute reduction
core attribute
swapping technology