摘要
提出一种线性/非线性时间序列模型的一般表达式(GNAR),论述其线性和非线性特性.对3种典型的非线性、非平稳时间序列进行试验及应用研究.将样本数据分成训练集和测试集,在训练集上建立GNAR模型,采用最小二乘方法以及结合预测误差的修正AIC准则实现其参数估计和模型定阶.在测试集上进行预测,进而验证模型.结果表明该模型对3组数据跟踪性能良好,预测预报精度优于传统时序模型,因此该模型有良好的适应性和有效性,能应用于工程实际.
A general expression for linear and nonlinear auto-regressive time series models（GNAR）is proposed and its linear and nonlinear characteristics are discussed.The model is verified by three typical data series which are divided into training and test sets.The GNAR model is established on the training set with the least square method to realize the parameter estimation and an information criterion integrated with the prediction error to determine the model order.And the fitting adequacy and the prediction error are checked on the test set.The model simulation and experiments show that the proposed GNAR model can accurately trace the dynamic characteristics of the nonlinear data and its modeling and prediction accuracy is superior to the traditional time series models.Therefore the GNAR model is flexible and effective and can be applied to the engineering.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第6期1077-1080,共4页
Journal of Southeast University：Natural Science Edition
基金
国家高技术研究发展计划(863计划)资助项目(2006AA040202)
江苏省科技成果转化资助项目(BA2006068).
关键词
线性/非线性系统
时间序列模型
系统辨识
linear and nonlinear system
time series model
system identification
