GRNN神经网络在粮食产量预测中的应用  

Application of GRNN Neural Network in Grain Yield Prediction

作　　者：薛培 

机构地区：[1]天津工业大学,电气工程与自动化学院,天津

出　　处：《数据挖掘》2020年第4期247-253,共7页Hans Journal of Data Mining

摘　　要：我国作为一个粮食种植大国,对粮食产量预测问题进行研究尤为重要。尽管我国的粮食产量每年都在稳定增长,但影响粮食产量变化的因素依然存在,例如:粮食播种面积、有效灌溉面积、受灾面积、农用化肥施用折纯量和从业人数等等,具有极其复杂的非线性关系。为了提高粮食产量的预测精度,通过BP神经网络和GRNN (广义回归神经网络)两种方法对比,并根据影响预测粮食产量变化的五种因素,分析神经网络模型、学习方法和网络结构等,对神经网络的参数进行优化,建立粮食产量的预测模型来精确预测粮食产量。本文在国家统计局统计1995~2019年全国粮食总产量及影响因素等数据基础上,建立BP神经网络与GRNN的仿真模型。预测结果表明:相对于BP神经网络,GRNN预测精度更高,速度更快,模型更稳定,可以很好地用于粮食生产预测。As a large country of grain cultivation, it is very important to study the problem of grain yield pre-diction in China. Although our country’s grain production is increasing steadily every year, the fac-tors affecting the change of grain production still exist, such as: grain planting area, effective irriga-tion area, affected area, fertilizer application amount and number of employees, etc., which have extremely complex nonlinear relationship. To improve the prediction accuracy of grain yield, the BP neural network and GRNN (generalized regression neural network) were compared, and the neural network model, learning method and network structure were analyzed according to the five factors affecting the prediction of grain yield. By optimizing the parameters of the network, the prediction model of grain yield is established to accurately predict grain yield. This paper establishes a simu-lation model of BP neural network and GRNN based on the statistics of the National Bureau of Sta-tistics 1995~2019. The prediction results show that compared with the BP neural network, the GRNN prediction accuracy is higher, the speed is faster, and the model is more stable, which can be well used in the prediction of grain production.

关 键 词：粮食产量 预测 BP神经网络 广义回归神经网络 

分 类 号：F32[经济管理—产业经济]