基于NARX神经网络的农产品价格时间序列预测方法研究  

Time Series of Agricultural Product Price Forecast Based on Nonlinear Auto-regression with External Input Neural Network

作　　者：彭琳[1,2] 林明[1] 

机构地区：[1]云南农业大学云南省高校农业信息技术重点实验室,昆明650201 [2]上海大学计算机工程与科学学院,上海200072

出　　处：《农机化研究》2013年第11期18-21,共4页Journal of Agricultural Mechanization Research

基　　金：国家自然科学基金项目(31260292);云南省自然基金项目(2008ZC050M)

摘　　要：针对传统时间序列预测方法在非线性时间序列预测上的不足,引入了非线性有源自回归神经网络(NARX),建立了基于非线性有源自回归神经网络农产品价格时间序列预测模型。该模型利用核函数对农产品价格时间序列进行数据变换;再用统计分析方法对模型性能进行评价、分析,进而对模型性能进行优化。实验结果表明:非线性有源自回归神经网络较传统时间序列预测模型,对非线性时间序列预测有更好的适应性和更高的预测精度。This paper introduced a nonlinear auto-regression with external input neural network（NARX） to solve the problem that traditional time series forecast method is impossible to analyze the nonlinear time series matters.A time series prediction model for agricultural products was applied in this paper based on NARX.Firstly,a kernel function was applied into the new prediction model to realize data transformation from time series of agricultural products.Secondly,statistical analysis was applied to evaluate the functions of the new prediction model and then to optimize the prediction model.The results showed the new model based on NARX has better adaptability and can predict more accurate than the traditional time series prediction model.

关 键 词：非线性有源自回归 神经网络 时间序列 统计分析 农产品价格 

分 类 号：S126[农业科学—农业基础科学]