关于粗糙集属性约简的进化算法研究和应用  

The Study and Application about the EA of Rough Set Attribute Reduction

作　　者：于冰[1] 阎保平[1] 

机构地区：[1]中国科学院计算机网络信息中心,北京100080

出　　处：《微电子学与计算机》2005年第3期189-194,共6页Microelectronics & Computer

摘　　要：粗糙集理论是一种新的处理模糊和不确定知识的数学工具.属性约简是粗糙集理论研究中的重要内容之一,本文提出了用其构建科学、合理、简洁有效的科研项目评审指标体系,为科研智能管理提供新的解决方法。本文中提出一种基于进化算法的知识相对约简算法。通过在知识表达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并通过决策属性支持度定义条件属性对决策属性的相对重要性以,,此作为启发式信息求出相对核,并将相对核加入进化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。Rough set theory is a new mathematical tool to deal with vagueness and uncertainty. Reduction of attribute is one of the important topics in the research on rough set theory. This paper deals with how to make a more efficient construct scientific, reasonable, compact index-system about evaluation of scientific research project and give a new method of solving the problem about management of scientific research. A kind of knowledge relative reduction Algorithm was proposed in this paper. With decision attribute support degree applied in knowledge express system, the support degree of the knowledge supplied by condition attribute for the whole decision was described and relative importance degree and relative core was obtained and relative core was obtained and as initial population in EA in order to accelerate convergence. Punishing function was used in fitness function to assuring reduction have fewer attributes and stronger support and search effect is very good. The practical results showed that the approach was effective in solving knowledge reduction.

关 键 词：可变精度粗糙集 知识获取 智能知识处理 智能计算 进化算法 数据挖掘 知识发现 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]