基于一致性度量的属性约简的研究  

Research of Attribute Reduction Based on Consistency Measure

作　　者：陈堃[1] 李心科[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《计算机技术与发展》2008年第10期64-67,共4页Computer Technology and Development

基　　金：安徽省科技计划项目(0012021A)

摘　　要：粗糙集理论作为一种处理不精确和不一致数据的数学工具被广泛应用于特征子集选择和属性约简中。在大多数现存的算法中,属性依赖度被用来度量特征子集的重要性,而依赖度在处理不一致信息系统时会出现找不到任何特征子集的问题。文中讨论了使用属性依赖性作为度量的缺点和不足,引入一种一致性度量,分析了其和依赖性之间的关系,重新定义了信息系统的多余属性和约简的概念,并构造了基于一致性度量的前向贪婪搜索算法。通过UCI数据集合验证了算法能够有效地处理不一致信息系统。As a new mathematic method which analyses and treats the inexact, incomplete information and knowledge, rough sets has been widely used in feature subset selection and attribute reduction. In most of the existing algorithms, the dependency measure is employed to evaluate the quality of a feature subset. In this paper, discuss the disadvantages and problems of using dependency, and introduce the consistency measure to deal with the problems. The relationship between dependency and consistency is analyzed. Redefine the redundancy and reduction of rough sets, and construct a greed search algorithm to find the reduction based on consistency. The experimental results with UCI data set show that the new algorithm is effective and efficient.

关 键 词：粗糙集 属性重要度 一致性 依赖度 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]