摘要
为了提高滑动平均模型ARMA(p,q)的预测精度,将模型参数估计转化为无约束优化问题,结合非线性规划中的共轭方向思想,提出一种改进的共轭梯度法:即结合不同共轭梯度法的优势,提出新的参数标量和搜索方向迭代公式,并证明该方法的全局收敛性.用此改进方法来修正原始ARMA(p,q)模型的参数估计值,给出数值算例,进一步验证所提方法的有效性.
For improving the prediction accuracy of auto regression moving average model,the parameter estimation was translated into unconstrained optimization problems in this paper.A improved conjugate gradient method was put forward based on the conjugate direction ideas in nonlinear programming.The new recurrence formula of scalar parameters and search direction was put forward combined with the goodness of different conjugated gradient method,and the global convergence of this method was proved.The parameters of original ARMA（p,q） were revised by using the method.The numerical example was given,and the results showed the availability of this method.
出处
《兰州理工大学学报》
CAS
北大核心
2014年第1期144-147,共4页
Journal of Lanzhou University of Technology
基金
国家自然科学基金(51175448)
河北省教育厅基金(2009159)
关键词
ARMA(p
q)模型
共轭梯度法
全局收敛
参数估计
非线性规划
ARMA（p, q） model
conjugate gradient method
global convergence
parameter estimation
nonlinear programming
