高等学校化学学报 ›› 1999, Vol. 20 ›› Issue (9): 1371.

• 论文 • 上一篇    下一篇

遗传算法用于变量筛选

章元, 朱尔一, 庄峙厦, 王小如   

  1. 厦门大学化学系, 教育部材料和生命过程分析科学开放研究实验室, 厦门 361005
  • 收稿日期:1998-11-12 出版日期:1999-09-24 发布日期:1999-09-24
  • 通讯作者: 朱尔一
  • 作者简介:章元,男,26岁,硕士研究生.
  • 基金资助:

    教育部留学回国人员科研启动基金;福建省自然科学基金

Variable Selection by Genetic Algorithms

ZHANG Yuan, ZHU Er-Yi, ZHUANG Zhi-Xia, WANG Xiao-Ru   

  1. Department of Chemistry, the Key Laboratory of Analytical Science for Material and Life Chemistry of MOE, Xiamen University, Xiamen 361005, China
  • Received:1998-11-12 Online:1999-09-24 Published:1999-09-24

摘要: 利用遗传算法的优越搜索寻优特性,结合有序Gram-Schmidt正文化及PLS算法可得到预报能力较强的模型,即PRESS(预报残差平方和)值较低的模型.该法可用于处理构效关系及人发微量元素与性别关系问题,并与正交递归选择法及逐步回归正向选择法进行比较,结果良好.

关键词: 遗传算法, 有序Gram-Schmidt正文化, PLS回归, 正交递归选择法, 变量筛选

Abstract: The model with a higher predictive ability can be obtained or the lower PRESS statistic values of the model can be achieved by use of coinbining the genetic algorithms with Gram-Schmidt orthogonalization, PLS method.The comparison is made among forward selection method in the stepwise regression and orthogonalization recurrence selection method as well as genetic algorithms in dealing with the examples of QSAR (quantitative structure activity relationship) and the problem of relationship between the trace elements in human hair and sex.

Key words: Genetic algorithms, Ordered Gram-Schmidt orthogonalization, PLS regression, Orthogonalization recurrence selection, Variable selection

中图分类号: 

TrendMD: