A Study of Crop Yield Prediction Based on Backpropagation Artificial Neural Network (BP- ANN)

Under the background of food security, the research of crop growth models has attracted much attention. There is an urgent need for simplified crop yield forecasting methods based on environmental factors while maintaining as much accuracy as possible. Firstly, 18 years of data were collected from a region, and the data were comprehensively analyzed and sorted into data including the yield of wheat, sorghum, and millet in each of the subordinate districts, as well as the seasonal climatic environment and soil type of each district. Secondly, a backpropagation artificial neural network (BP-ANN) model was established to predict the crop yield according to the environmental factors. Then, Python was used for simulation, and the conclusion was as follows: the optimal values of mean square error (MSE) for the training and validation set were 0.2555 and 0.2497 respectively, which verified the validity and accuracy of the model. Finally, compared with the random forest (RF) method and support vector machine (SVM) method, it was found that the effect of the BP-ANN method was better. In addition, the Spearman's correlation coefficient among different variables was analyzed, and it was found that there was little correlation between each variable and yield, which verified that the BP-ANN model was good at capturing complex data characteristics and variable relationships, and this method provided an effective reference for standardized agricultural planting and crop yield prediction.