| In recent years,the issue of food security has aroused widespread concern around the world.As the world’s most populous country,ensuring adequate food production is a top priority for China’s national development.The report of the 20 th National Congress of the Communist Party of China points out that "comprehensively consolidating the foundation of food security" is the basis for rapid economic development,the guarantee of national and political security,and the condition for social harmony and stability.In order to ensure adequate food supply for the Chinese population,better planning of grain production,scientific and effective prediction of grain production and reasonable arrangement of grain production are key means to solve the food security problem.This paper focuses on the yield of rice in Guangxi and the related factors affecting rice yield,and studies the prediction of grain yield using Bayesian LightGBM model.LightGBM model is a model with excellent neutral performance that has been applied to regression prediction in recent years,mainly featuring fast training speed and less memory consumption.It has been widely used in the fields of machine learning and data mining,but currently no one has applied it to agricultural science.Due to the impact of multiple factors on grain production,in order to highlight the advantages of LightGBM model in dealing with multiple influencing factors,this article compares the 18 dimensional feature input models after data preprocessing.Experiments have shown that LightGBM has higher prediction accuracy and faster speed.In order to further improve the prediction accuracy,based on the LightGBM model,the loss function is modified to the Huber loss function,and a Bayesian optimization algorithm is introduced to determine the optimal hyperparameter combination and input it into the model.A Bayesian-LightGBM model is proposed.The experimental results show that the MAE of the prediction model based on linear regression is 1.255,the MAE of the prediction model based on decision tree is0.426,the MAE of the prediction model based on random forest is 0.315,and the MAE of the prediction model based on Bayesian-LightGBM is 0.049.Compared to other prediction models,Bayesian-LightGBM can more effectively predict grain yield,and the prediction accuracy is better than traditional machine learning models. |
