基于时间序列与RBF的农产品市场价格短期预测模型

Short-term forecast of agricultural products price based on time series and RBF

为协助生产者更好地播种和收获,有效把握国内农产品价格波动规律,提高农产品价格预测精度。本文构建了基于农产品价格时间序列组合预测方法与RBF神经网络的集成预测模型,以ARCH、Holt-Winters无季节模型时间序列组合预测方法揭示农产品价格序列线性特征,以RBF神经网络揭示农产品价格非线性变动规律,并以1997—2011年全国农产品集贸市场大豆月度价格走势数据为例进行实验验证。研究结果显示,基于农产品价格时间序列组合预测方法与RBF神经网络的集成预测模型精度高于时间序列组合预测方法或RBF神经网络模型,是一种有效的农产品价格预测模型。

An integrated prediction model is proposed based on time series model and RBF, forecasting the trend of agricultural products price, for the purpose of assisting producers in seeding and harvesting. The time series prediction model ARCH and Holt-Winters no season are weighted composited, featuring the linear characteristic of price series. While RBF is added afterward, featuring the nonlinearity characteristic. A test on monthly price series of soybean covering 1997 to 2011 is carried out. The result that the accuracy of the integrated forecasting model is higher than either combined time series model or RBF proves the effectiveness of this model.