利用自回归模型的平稳时序数据快速辨识算法  

Fast identification algorithm for stationary time series data using auto-regressive model

作　　者：黄雄波[1] 胡永健[2] Huang Xiongbo;Hu Yongjian(Dept.of Electronic&Information Engineering,Foshan Professional Technical College,Foshan Guangdong 528000,China;School of Electronic&Information Engineering,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]佛山职业技术学院电子信息系,广东佛山528000 [2]华南理工大学电子与信息学院,广州510641

出　　处：《计算机应用研究》2018年第9期2643-2647,共5页Application Research of Computers

基　　金：广东省自然科学基金团队项目(9351064101000003);广东省应用型科技研发专项基金资助项目(2015B010130003);广东省科技计划工业攻关项目(2011B010200031);佛山职业技术学院校级重点科研项目(2015KY006);佛山职业技术学院横向重点资助项目(H201813)

摘　　要：由于自回归模型的参数估计可归结为求解一个线性方程组的问题,所以其在平稳时序数据的辨识过程中具有广泛的应用场合。提出了一种基于自回归模型的快速辨识算法,以递推的方式对平稳时序数据自相关函数矩阵的秩的下界值进行估计,再以该估计值作为自回归模型的起始阶数对系统进行依次的递阶辨识。最后,基于F检验对相邻阶次的拟合误差的变化趋势进行显著性检验,并以检验结果作为算法的结束条件。新算法在保证较高辨识精度的条件下,其计算效能及辨识精度的稳定性均优于现有的自回归模型辨识算法,实验结果验证了新算法的有效性和先进性。Because the parameter estimation of auto-regressive model can be reduced to the solution of a linear equation group,it has a wide range of applications in the identification of stationary time series data.This paper proposed a fast identification algorithm based on auto-regression model firstly,bound by recursion on the stationary time series data auto-correlation matrix rank estimation,then,estimated value as the starting order regression model in turn hierarchical identification,finally,the fitting error of adjacent F checked the order of significance test based on the test results,and as the end condition of algorithm.With the auto-regressive model identification algorithm compared with the existing,because the new algorithm avoided in the hierarchical identification from the first order,and thus it obtained better computing performance.The experimental results verify the effectiveness and superiority of the new algorithm.

关 键 词：平稳时序数据 自回归模型 递阶辨识 自相关函数 

分 类 号：TP311.11[自动化与计算机技术—计算机软件与理论] TP301.6[自动化与计算机技术—计算机科学与技术]