RBF神经网络的粮食产量预测

摘要

The grain production prediction accuracy is an important problem. There is some nonlinear relation between the grain yield and its influencing factors, and the traditional grain production forecast method cannot capture the nonlinear rule, which causes low prediction accuracy. In order to improve the accuracy of predicting grain production , RBF neural network is introduced into the grain production of the prediction process. This method adopts RBF neural network to grain yield prediction model by using the nonlinear fixed order by determining the order number of data sample reconstruction, and builds a multiple input and single output of grain production samples. Then genetic algorithm is used to optimized the RBF neural network model parameters, and the optimal food production forecast model is established. Simulation experiments are carried out with the Chinese grain production. The results show that compared with ARIMA model, sliding average model and support vector machine prediction model, the prediction accuracy of RBF neural network model is higher than the other model and the speed is faster. The method provides a new way for the grain production prediction.%研究粮食产量准确预测问题,粮食产量变化多种因素综合结果,针对各因素间具有十分复杂的非线性关系,传统预测方法无法反映粮食产量非线性变化规律,导致粮食产量预测精度低.为了提高粮食产量预测精度,提出一种RBF神经网络的粮食产量预测方法.通过采用非线性能力强的RBF神经网络对粮食产量数据进行非线性定阶,通过最优阶数对粮食产量模型进行重构,然后利用遗传算法对RBF神经网络参数进行优化,获得粮食产量最优预测模型,最后对粮食产量进行预测.通过对1 978 - 2008年中国粮食产量进行仿真,实验结果表明,相对于ARIMA、滑动平均和支持向量机等预测模型,RBF神经网络预测精度高,速度快,证明改进方法为粮食产量预测提供了一种新的途径.