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.
出处
《农机化研究》
北大核心
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)
