Prediction of rice yield based on LSTM long short term memory network

Abstract

The prediction of rice yield is of great significance for improving rice yield and preventing disasters. At present, there is no exact and effective solution to the problem of rice yield prediction, and there are many factors affecting the rice yield, which may lead to large errors in the prediction results. Based on the rice yield data of 81 counties in Guangxi Province of China for three years, a long short time memory network (LSTM) model was established and used to predict rice yield under the influence of meteorological factors. The results show that the performance of the recurrent neural network with LSTM structure is better than that of the standard recurrent neural network, which solves the problems of gradient dispersion and gradient explosion of RNN model, and enables the model to learn long-term laws. At the same time, the traditional machine learning model is established to compare with LSTM model. The neural network model can better mine the information hidden in the data and make full use of the rules in the data. Combined with meteorological factors, the test shows that LSTM model has higher prediction accuracy. It provides a reference for rice yield prediction.