一种具有记忆功能的遗传算法属性约简方法  

Method of attributes reduction based on GA with memory function

作　　者：张忠林[1] 张军[1] 米伟[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,兰州730070

出　　处：《计算机应用研究》2010年第1期96-98,共3页Application Research of Computers

基　　金：兰州交通大学"青蓝"人才工程基金资助项目(QL-05-10A)

摘　　要：通过对粗糙集属性约简算法现状的分析,针对遗传算法求取属性约简中存在的问题,提出一种基于具有记忆功能遗传算法粗糙集属性约简的新算法。该算法通过在编码过程引入一个标志位,使遗传算子具有一定的记忆功能,从而分别对两个子群进行搜索,同时只对其中一个子群采用最佳个体保存法,提高了收敛的概率,保证了最终结果的正确性。实验表明,该算法在收敛性和正确性方面优于基于传统遗传算法的属性约简算法。Through the analysis of present algorism of attributes reduction for rough sets, focusing on the problems that the algorism of attributes reduction based on GA has, this paper proposed a new algorithm of attributes reduction for rough sets based on a GA which had memory function. This algorithm made the GA have certain memory function by importing a flag bit, consequently, it could search in the two subgroups respectively. At the same time, it was only one subgroup that adopted the elitist model. All these could increase the probability of convergence and ensure correctness of the result. The experiment shows that the new algorithm is more effective than the algorithm of attributes reduction based on traditional GA.

关 键 词：粗糙集 遗传算法 属性约简 核 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]