基于遗传算法的特征子集选择  

Feature Subset Selection Based on Genetic Algorithm

作　　者：刘勇国[1] 李学明[1] 张伟[1,2] 彭军[1,3] 廖晓峰[1] 吴中福[1] 

机构地区：[1]重庆大学计算机科学与工程学院 [2]重庆教育学院计算机与现代教育技术系,重庆400067 [3]重庆工业高等专科学校计算机系,重庆400050

出　　处：《计算机工程》2003年第6期19-20,50,共3页Computer Engineering

基　　金：重庆市科技计划基金资助项目

摘　　要：数据挖掘对象是大型数据库中的海量数据,而数据库中记录包含众多属性,由于其中存在的冗余和不相关属性降低了数据挖掘性能,增加了算法复杂性,因此,特征子集选择问题成为数据挖掘领域中的重要研究课题。该文根据过滤法思想,提出了基于遗传算法的特征子集选择算法,实验证明该算法获得了良好的收敛性和稳定性。Data mining aims at huge data in the very large database. Records in this database include many redundant and irrelated features, which degrade the data mining performance and increase the algorithm complexity. So, feature subset selection becomes one of the important research issues in the field of data mining. In this paper, an algorithm of feature subset selection based on genetic algorithm is proposed according to the filter approach. It is proved by experimental results that the convergence and the stability of this algorithm are adequately achieved.

关 键 词：数据库 数据挖掘 遗传算法 特征子集选择算法 模拟退火算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]