A Meteorological Data Prediction Algorithm Based on Recurrent SwinTransformer

Meteorological data prediction belongs to the category of spatio-temporal data prediction, which is usually based on the radar image sequence of the past period to predict the radar image sequence of the future period. This problem is usually subject to multi-dimensional data changes, which puts forward higher requirements for feature extraction and long-term memory ability of the model. Accordingly, we propose a meteorological data prediction algorithm, Recurrent SwinTransformer, based on SwinTransformer and RNN (Recurrent Neural Network). we introduce a recursive structure based on sliding window mechanism to solve the long-term memory problem of RNN. Then, a crawler is used to collect the meteorological radar data of southwest China in the past year. Aiming at this data set, we make comparative experiments between SwinLSTM algorithm and the proposed algorithm. The experimental results show that the algorithm in this paper has achieved better performance in the forecast of meteorological radar data.