基于粗糙集的启发式属性约简算法  

Heuristic algorithm of attribute reduction in condition entropy

作　　者：施化吉[1] 秦川[1] 陈海军[2] 李星毅[1] 

机构地区：[1]江苏大学计算机学院,江苏镇江212013 [2]太原铁路局,山西太原030013

出　　处：《计算机工程与设计》2008年第19期5014-5015,5050,共3页Computer Engineering and Design

基　　金：国家火炬计划基金项目(2004EB33006)

摘　　要：对现有启发式属性约简算法进行分析,通过实例说明一般启发式算法求得的相对约简有冗余属性存在的问题。针对这一不足,利用粗糙集理论中的条件熵作为启发信息,来缩小搜索空间,并在算法中加入消除冗余属性的二次约简过程,得到一种改进的启发式属性约简算法。提供了实例分析,验证了该改进算法具有较好的约简效果。The present heuristic algorithm of attribute reduction is analyzed. The problem that the redundancy attributes still exist in relative reduction is shown from cases. To this issue, the condition entropy is used as heuristic information for reducing the searching room. And a process of reducing again is used in the improved algorithm in order to eliminate redundancy attributes. So an improved heuristic algorithm of attribute reduction is presented. The exarnple is presented and the result validates the improved algorithm with good reduction effect.

关 键 词：条件熵 粗糙集 属性约简 启发式算法 核 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]