【摘要】 从函数逼近和系统辨识两个方面推导了非线性自回归时序模型(GNAR模型)的物理结构,通过公式推导及仿真数据研究GNAR模型与确定性实函数、经典时序模型和混沌序列的关系,明确GNAR模型对系统逼近的机理.以Lorenz系统输出的混沌序列和现代经典时序-太阳黑子序列为算例进行数据实验,证明了GNAR模型在建模和预测方面的优越性.更多还原

【Abstract】 The general expression for nonlinear autoregressive model(GNAR model) was derived based on the functional approximation or system identification.And the correlation was improved between GNAR model and real functions,classical time-series models or chaotic sequences through formula derivation and simulation data experiments to determine the approximation mechanism of GNAR model.Finally the fitting experiments of chaotic sequence obtained from Lorenz system and the modern time series datasunspots show the superiority of GNAR model in modeling and forecasting.更多还原