基于BP神经网络的河北省耕地生产力预测  

The Forecast of the Cultivated Land of Hebei Province Based on BP Neural Network

作　　者：刘磊[1] 刘瑞卿[1] 石剑[1] 李新旺 张路路 霍习良[1] 

机构地区：[1]河北农业大学国土资源学院,河北保定071001 [2]河北省水土保持工作总站,石家庄050021 [3]河北省土地学会,石家庄050091

出　　处：《农机化研究》2012年第5期26-29,共4页Journal of Agricultural Mechanization Research

基　　金：国家重点基础研究发展规划项目(2005CB121107)

摘　　要：鉴于BP网络在处理非线性复杂系统的优势,以河北省为研究对象,构建一个9-5-1结构的BP神经网络预测模型,将1987-2005年的相关数据作为模型的训练样本,以2006年的粮价政策、农资投入量和农民收入等数据作为网络的预测输入,对该年的河北省粮食单产进行预测。结果表明,BP神经网络预测结果与实际粮食单产的相对误差为0.86%,预测精度优于传统的多元回归统计模型。Since BP network has the advantage in dealing with nonlinear complex systems,a BP neural network predictor model of 9-5-1 structure was built in this paper.It took Hebei province for example and made the relevant data of 1987-2005 as the training samples of the model.The forecast of grain yield per unit area of Hebei province would be known by taking the policy of grain price,the count of agriculture inputs and income for farmers as the prediction input of network.The results showed that the relative error between the consequence forecasted by BP neural network and the actual grain yield per unit area was 0.86%.BP neural network has a higher prediction accuracy than regression model of multivariate statistics.

关 键 词：粮食单产 耕地生产力 BP神经网络 河北省 

分 类 号：F307.11[经济管理—产业经济]