摘要
把数据拟合建模看作是模型结构和参数的优化搜索过程,将遗传规划和最小二乘法结合起来共同实现数据拟合,该方法是先用遗传规划随机生成待拟合函数的结构,再利用最小二乘法确定参数,然后耦合后生成待拟合的函数,实现模型结构和参数的共同识别.数值仿真表明,该方法得到的函数比传统方法得到的函数的拟合精度要高,且计算精度高.
Data fitting modeling can be considered as optimal search processes of model structures and model parameters. A new genetic evolutionary method, combining genetic programming and least square method is proposed. In this method, the construction of fitting function is got randomly by genetic programming, and parameters are determined by the least square method, where the structures and parameters coexist and interact. The experimental results show that the function got by this method is not only more accurate than traditional one, but also more ideal than some intelligence models.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2006年第B07期527-530,共4页
Journal of Harbin Engineering University
基金
国家自然科学基金(60461001)
广西自然科学基金(0542048)和广西民族大学重大项目研究基金资助项目.
关键词
遗传规划
最小二乘法
数据拟合
均方差
genetic programming
least square method
data fitting
square deviation