Research on grain yield prediction based on Bayesian-LightGBM model

At present, the grain yield prediction models, such as the grey relational model, generally have problems such as slow training speed and low prediction accuracy. In order to solve the above problems, this paper is based on the Lightweight Gradient Boosting Machine（LightGBM） model, and its loss function is modified to a Huber loss function, and a Bayesian optimization algorithm is introduced to determine the optimal hyperparameter combination and input into the model. Simulation experiments were carried out on the data sets of early and late rice yields and 16 grain yield influencing factors in Guangxi. The results showed that the average absolute error of the prediction model based on linear regression was 1.255, the average absolute error of the prediction model based on decision tree was 0.426, the average absolute error of the prediction model based on random forest was 0.315, and the average absolute error of the prediction model based on Bayesian LightGBM was 0.049. Compared with other prediction models, Bayesian LightGBM grain yield prediction model can realize grain yield prediction more effectively, with higher prediction accuracy.