基于互信息变化率的属性约简算法  

Algorithm for Attribute Reduction based on Mutual-Information

作　　者：陈帆[1] 林和[1] 石周[1] 崔伟男[1] 

机构地区：[1]兰州大学信息科学与工程学院,甘肃兰州730000

出　　处：《信息网络安全》2012年第7期40-42,共3页Netinfo Security

摘　　要：属性约简是粗糙集理论中的基本内容之一,但获得决策系统的最小相对约简是一个NP问题。文章从信息论的角度,结合微积分中的变化率思想,定义了一种基于互信息变化率的属性重要性度量方法,然后在此基础上提出了两种相对属性约简的启发式算法。最后,通过实验结果证明该方法能有效地对决策系统进行约简。Attribute reduction is a fundamental issue in the theory of rough sets. But it has been proved that computing the minimum reduction of decision table is a NP-hard problem. In the paper, according to the concept of change rate, a new attribute importance measure method is defined from the viewpoint of information theory, and two algorithms for Attribute Reduction based on mutual-information change rate are proposed. At last, the experimental results show that the algorithms can reduce the decision system effectively.

关 键 词：粗糙集 属性约简 互信息 变化率 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]