一种基于一致性准则的属性约简算法.docVIP

一种基于一致性准则的属性约简算法 杨明 （南京师范大学计算机科学与技术学院,南京 210097） 摘要：粗糙集方法提供了一种新的处理不精确、不完全与不相容知识的数学工具. 属性约简是粗糙集理论的重要研究内容之一，已有的大多数属性约简算法主要针对离散值属性的，面向连续值属性的约简报道较少.为此，在引入新的一致性定义后，提出一种新的基于一致性准则的属性约简模型，该模型可针对离散或连续值属性进行有效的约简,提出了一种基于一致性准则的属性约简算法，该算法可有效进行连续值属性的约简错分对象数的控制可有效增强属性约简的有效性.理论分析和实验表明提出的算法是有效可行的.粗糙集；属性约简；一致性准则A novel algorithm for attribute reduction based on consistency criterion YANG Ming ( School of Computer Science and Technology, Nanjing Normal University,Nanjing 210046 ) Abstract: Rough set theory is a new mathematical tool to deal with imprecise, incomplete and inconsistent data. Attribute reduction is one of important parts researched in rough set theory. Many existing algorithms mainly aim at the reduction of discrete-valued attributes, very little work has been done for attribute reduction aiming to continuous-valued attributes. Therefore, in this paper, after introducing a new definition on consistency of objects, we introduce a novel model based on consistency criterion for attribute reduction. The newly designed model is very suitable for the decision table with discrete-valued or continuous-valued attributes, and an extension of the classical rough set model. Based on this model, a novel algorithm for attribute reduction based on consistency criterion is proposed. This algorithm can effectively obtain an attribute reduction for the decision table with continuous-valued attributes, and meanwhile the effectiveness of the attribute subset obtained by the new model can be enhanced by controlling the number of the misclassified or consistent objects. Theoretical analysis and experiments shows that the algorithm of this paper is efficient and feasible. Key words: (rough set; attributes reduction; consistency criterion 1 引言 波兰数学家Z.Pawlak 80年代初提出的Rough Set(RS,粗糙集)是一种新的处理不精确、不完全与不相容知识的数学理论[1],近年来该理论在机器学习、数据挖掘及模式识别等多个领域得到了广泛的应用[2].在粗糙集理论中，属性约简是重要研究内容之一，也是知识获取的关键步骤.现有的属性约简大体上可分为[5-9]、基于正区域的属性约简算法[10-11]、基于启发式的属性约简算法[12-13]三种，但这些基于经典Rough集模型的算法都是针对离散值属性，不适合连续值属性的约简.目前，在粗