Abstract:In order to improve the accuracy of sea level change prediction, we propose combining CEEMDAN with LSTM to get the CEEMDAN-LSTM method adopting the idea of decomposition-prediction-recombination. The results show that the combined CEEMDAN-LSTM method outperformed the direct LSTM prediction(MAE=16.87 mm,RMSE=21.51 mm) and the existing combined EEMD-BP neural network method(MAE=10.4 mm,RMSE=15.44 mm) with lowest prediction error (MAE=8.89 mm,RMSE=11.34 mm).