摘要
针对约简属性组合的爆炸问题,将RS属性核参数作为先验信息的免疫疫苗引入抗体编码,概率性对种群接种疫苗.将属性集合的分类近似标准作为抗体适应度,通过在免疫克隆选择过程中引入聚类竞争机制,提高抗体群分布的多样性及亲和力成熟,从而获得多个属性约简及最小约简的平衡.实验结果表明,这种粗糙集属性约简方法对于多维条件属性集是快速且有效的.
Aiming at the problem of large scales of attribute reduction, a prior knowledge of attribute kernel as bacterins is introduced to antibody coding and the population is vaccinated in a stochastic way. The classification approximation quality is taken as the antibody affinity , then a cluster and competition mechanism is applied in clonal selection process to enhance the diversity of antibody and affinity maturation. Minimum reductions and more reductions of the rough sets are obtained. Experimental results show that the approach＇is effective and quick in solving attribute reduction, and has a remarkable quality of the global convergence reliability and convergence velocity.
出处
《控制与决策》
EI
CSCD
北大核心
2008年第5期497-502,共6页
Control and Decision
基金
国家自然科学基金重点项目(60634020)
教育部博士点基金项目(20060532026)
关键词
克隆选择
粗糙集
属性约简
数据挖掘
Clonal selection
Rough sets
Attribute reduction
Data mining
