非线性自回归模型

On a Class of Nonlmear Autoregressive Models

对由Volterra泛函级数表示的一般非线性模型做一定的限制,得到一类简单的由线性系统和无记忆非线性系组组成的非线性Block-oriented模型。本文假定过程是可分离的,根据非线性投影原理,得到修正后的Wiener-Hopf方程,其中修正系数反映了过程的高阶统计信息.Block-oriented模型中的线性部分和非线性部分的参数可分开来辩识.本文还分析了对于可分离过程,采用非线性Block-oriented模型预测的预测性能相对于采用线性预测的性能改善程度.计算机模拟结果说明了文中的结论.

By imposing some restrains on the general nonlinear moder describea by Voiterra unct(?)na, a class of simple nonlinear block-oriented models formed with connecting linear system and nonmemer nonlinear system can be obtained. Considering nonlinear projection theorem and separab(?) for some processes, we can obtain the modified Wiener—Hopf equation with a coefficient involving some statist(?)cs(?) higher—order memonts. The identification procedure for linear and nonlinear part can be carrted out (?) vidually Analytical results indicate the performance improvement of nonlinear block—onentec predictiot over linear case for separable processes Some results of simulation studies are includea 'o illustra(?) the superiorty of such models to linear one in fit for the real data