摘要
提出一种基于遗传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 。
A kind of knowledge relative reduction Algorithm was proposed. 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 GA 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.
出处
《系统工程》
CSCD
北大核心
2003年第4期116-122,共7页
Systems Engineering
基金
国家自然科学基金资助项目 (70 1710 5 6)
国家重点科技攻关资助项目 (975 6 2 0 10 7)
关键词
人工智能
遗传算法
粗糙集理论
知识约简方法
决策属性
知识库
Rough Set Theory
Genetic Algorithms
Decision Attribute Support Degree
Relative Core
Relative Reduction
FitnessFunction
