改进的启发式LEM2快速规则获取算法  

Fast Algorithm for Rule Extraction Based on Heuristic LEM2

作　　者：纪霞[1] 李龙澍[1,2] 陈圣兵[1] 赵鹏[1,2] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039 [2]安徽大学计算机科学与技术学院,安徽合肥230039

出　　处：《小型微型计算机系统》2010年第11期2278-2281,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60273043)资助;安徽省科技攻关计划重大科技专项(08010201002)资助;安徽省高校拔尖人才基金项目(05025102)资助;安徽省自然科学基金项目(050420204)资助;安徽省教育厅重点项目(KJ2009A001Z)资助

摘　　要：LEM2算法是一种有效的规则提取算法,但是效率有待提高.针对效率问题,文章对LEM2算法作了一些改进:在计算属性值对的同时获取单条件属性的确定规则,并且可以去掉某些冗余属性;定义了决策分配势矩阵作为启发信息,简化了LEM2算法内部的多重循环.最后给出了两种算法的复杂度分析和仿真对比试验,结果表明改进算法能获取更为简洁的规则,并能有效的提高算法的效率.LEM2 is an effective rule extraction method except it＇s low efficiency. To boost LEM2 algorithm＇s efficiency, some improvements were taken in this paper. This paper directly extracted single attribute rules and deleted some redundant attributes synchronously with computing the attribute value pair. Besides that, it defined decision distribution force matrix as heuristic information to predigest the multiple circulation in LEM2. At last, the complexity analysis and contrast simulation experiment were given. Results show that the improved method can effectively boost the efficiency and make rules more laconic.

关 键 词：规则获取 决策分配势矩阵 启发信息 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]