基于遗传算法的不完备信息系统属性约简方法  

Attribute reduction based on GA under incomplete information system

作　　者：陶志[1] 刘庆拯[1] 李卫民[1] 

机构地区：[1]中国民航大学理学院,天津300300

出　　处：《系统工程与电子技术》2007年第9期1484-1487,共4页Systems Engineering and Electronics

基　　金：国家自然科学基金委员会与中国民用航空总局联合资助项目(60672178);中国民航大学博士启动基金资助项目(05qd02s)

摘　　要：提出了一种基于遗传算法的属性相对约简算法。通过在不完备信息系统中引入决策属性支持度的概念,来描述由条件属性所提供的知识对整体决策的支持程度,以此作为启发式信息求出相对核,并将相对核加入遗传算法的初始种群中以加快算法的收敛。同时,在适应值函数中引入惩罚函数,可以保证所求约简既含较少的属性又有较强的支持度,能够获得最佳的搜索效果。该算法通过实例分析,证明是求解属性约简问题的快速有效方法。A kind of attribute relative reduction algorithm is proposed. With decision attribute support degree applied in a incomplete information system, the support degree of the knowledge supplied by condition attribute for the whole decision is described and the relative importance degree and relative core are obtained as initial population in GA in order to accelerate convergence. A punishing function is led into the fitness function to assuring reduction having fewer attributes and stronger support and search effect is very good. The practical results show that the approach is effective in solving attribute reduction.

关 键 词：粗糙集理论 属性约简 遗传算法 不完备信息系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]