As a large country of grain cultivation, it is very important to study the problem of grain yield prediction in China. Although our country’s grain production is increasing steadily every year, the factors affecting the change of grain production still exist, such as: grain planting area, effective ir- rigation 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 simulation model of BP neural network and GRNN based on the statistics of the National Bureau of Statistics 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.
展开▼
