基于粗糙属性向量树的属性约简新方法  

A NEW METHOD FOR ATTRIBUTE REDUCTION BASED ON RAVT

作　　者：张迎春[1] 闫德勤[1] 张丹枫[2] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116081 [2]沈阳医学院护理学院,辽宁沈阳110034

出　　处：《计算机应用与软件》2010年第8期19-22,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(60372071)

摘　　要：在对区分能力大小研究的基础上建立了一个用于指导信息表的绝对属性约简的粗糙集模型,同时在对区分能力和分类能力二者关系深入研究的基础上提出了决策依赖区分精度新概念,该概念是用于指导决策表的、相对属性约简的一个新的判据。借助粗糙属性向量树提出了新的求全部属性约简的算法,通过理论分析说明了新算法的最坏时间复杂度低于经典的"基于差别矩阵求全部属性约简算法"以及它的改进算法。对比实验结果验证了该算法在运算效率上明显高于"基于差别矩阵求全部属性约简算法"的改进算法。A rough set model is established for supervising the absolute attribute reduction of information table on the basis of studying the capability of discernible ability. A novel conception, which is called decision-dependent discernibility precision, is proposed on the basis of thorough exploring the relationship between the abilities of discernibility and classification. This conception provides an important criterion to supervising the relative attribute reduction of decision table. A new algorithm for finding all the attribute reductions is proposed by means of rough attribute vector tree （RAVT）. By theoretical analyses, it has been illustrated that the worst time complexity of the new algorithm is less than that of the traditional algorithm of fining all the attribute reductions based on differential matrix and its improved algorithm. Comparative experiments result proves that this new algorithm saliently outperforms that improved algorithm on computational efficiency.

关 键 词：区分精度 决策依赖区分精度 粗糙属性向量树 差别矩阵 近似精度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]