基因表达式编程在组合预测建模中的应用  

Application of gene expression programming on combination forecasting modeling

作　　者：张大斌[1,2] 张景广[1] 彭森[1] 

机构地区：[1]华中师范大学信息管理系,武汉430079 [2]中国科学院数学与系统科学研究院,北京100190

出　　处：《系统工程理论与实践》2012年第3期568-573,共6页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(70971052);中国博士后基金(20080440539);华中师范大学中央高校基本科研业务费项目(CCNU09B01003)

摘　　要：介绍一种利用基因表达式编程的方法来自动生成非线性函数的组合预测模型,并进行误差估计分析,改变过去只依靠各个子方法的简单线性相加,不能很好地反映非线性真实世界的传统组合预测建模方法.通过对我国CPI的真实历史数据验证,验证结果表明:与传统的ARIMA,灰色GM(1,1),BP神经网络和线性组合预测四种方法对比,基因表达式编程建立的组合预测模型所预测的数据准确度明显提高.An application of gene expression progranmfing in combination forecasting modeling is proposed, which obviously improves tile traditional methods of linear combination that could not express nonlinear real world. Then, the estimating standard and forecasting standard error are calculated and analyzed. By using the actual historical data from CPI of China, the automatic generated combination forecasting model by using gene expression programming is established, and the result indicates that the accuracy calculated by this combination model is obviously much higher, compared with traditional methods such as ARIMA, GA（1,1）, BP neural network and linear combination forecasting.

关 键 词：组合预测 基因表达式编程 自动建模 函数挖掘 

分 类 号：F830[经济管理—金融学]