Research on Monthly Precipitation Prediction Based on Additive Ensemble Neural Network

Abstract

Monthly precipitation is a crucial indicator to measure the degree of drought and flood in a region, which directly reflects the local ecological changes and ecological environment. Considering the application of neural network technology in the field of artificial intelligence, in order to overcome the problem of the decrease of generalization ability caused by the increase of model complexity, an additive ensemble neural network model is proposed in this work, which integrates the extreme learning machine models with different kernel functions into a unified framework, and effectively utilizes the complementary information of each kernel function to improve the overall performance of the model. Finally, an integrated monthly precipitation prediction model is established based on the monthly precipitation data of Liuzhou from 1951 to 2020 in April. Experimental results show that this model has high precision and satisfactory stability, and has achieved decent results in the field of precipitation prediction, which provides a new theoretical guidance for neural network integration.