基于LOWESS的函数系数自回归模型(FAR)优化及应用  

Optimization of Functional-Coefficient Autoregressive Models(FAR)Based on LOWESS and Its Application

作　　者：苏理云 梁昌海 李凤兰[2] 赵胜利 宋江敏 SU Liyun;LIANG Changhai;LI Fenglan;ZHAO Shengli;SONG Jiangmin(School of Science,Chongqing University of Technology,Chongqing 400054,China;Library,Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]重庆理工大学理学院,重庆400054 [2]重庆理工大学图书馆,重庆400054

出　　处：《重庆理工大学学报（自然科学）》2020年第3期228-239,共12页Journal of Chongqing University of Technology：Natural Science

基　　金：重庆市科委基础研究与前沿探索项目(cstc2018jcyjAX0464)。

摘　　要：函数系数自回归模型(FAR)在非线性时序数据分析应用中,当样本值两端存在数据偏少或异常值的情况,模型回归系数的估计精度不高和稳定性差,引入局部加权散点平滑(LOWESS)方法,优化FAR模型的估计。首先采用重标极差分析法(R/S)计算Hurst指数,并判别序列特征,再建立FAR模型并估计回归系数,最后结合LOWESS平滑方法优化FAR模型,建立LOWESS-FAR模型(LW-FAR模型)。通过模拟实验和国泰基金收益率实证分析、预测,表明:LW-FAR模型克服了FAR模型存在异常值时的缺陷,并提高了模型的预测精度和稳定性。In the application of function-coefficient autoregressive models(FAR)to nonlinear time series data analysis,when there are few data or outliers at both ends of the sample values,the model has the problems of low accuracy and poor stability in estimating the regression coefficients.In this paper,the local weighted scattered smoothing(LOWESS)method is introduced to optimize the estimation of FAR model.Firstly,we calculate the Hurst exponent by using the method of rescaled range analysis(R/S),and distinguish the nonlinearity of the series.Then we establish the FAR model and estimate the regression coefficient.Finally,we apply the LOWESS smoothing method to optimize the FAR model,and the LOWESS-FAR model(LW-FAR model)is established.The simulation experiment and the empirical analysis and prediction of the fund rate of return show that the LW-FAR model overcomes the defects of the FAR model when there are outliers,and improves the prediction accuracy and stability of the model.

关 键 词：HURST指数 FAR模型 LOWESS平滑 预测精度 

分 类 号：O21[理学—概率论与数理统计]