摘要
属性约简是粗糙集理论研究的核心内容之一,是知识获取的关键步骤。针对大规模数据集,基于决策表差别矩阵属性约简不具备可操作性缺点;以及基于差别矩阵属性频度的约简算法没有考虑到差别矩阵元素中属性个数多少的缺陷。基于差别矩阵元素的基数越小,其属性越重要的思想,按照基数由小到大的顺序,利用矩阵中具有相同基数的矩阵元素的簇集中属性出现的频度,确定属性的重要度,提出一种快速搜索属性约简算法,能快速搜索到属性的最优或次优约简。实验结果表明算法是可行、有效的。
Attribute reduction is one of the cardinal contents of research for theory of rough sets, and is a key step of knowledge acquisition. According to the attribute that the number of attributes is less in a discernibility matrix element, a fast search attribute reduction algorithm is proposed by using the importance of attribute based on discernibility matrix of decision table and attribute's frequency in the union set of matrix elements. It avoids the unfeasibility of attribute reduction based on discernibility matrix in larger database and the flaw of attribute reduction algorithm based on attribute frequency of discernibility matrix because of the number of attributes unconsidered in discernibility matrix elements. It has proved to be effective by the result of experiment.
出处
《计算机仿真》
CSCD
2008年第12期118-121,共4页
Computer Simulation
基金
安徽省高校自然科学研究项目(KJ2008B125)
学院教研项目(8398)
关键词
粗糙集
差别矩阵
属性约简
决策表
Rough sets
Discernibility matrix
Attribute reduction
Decision table