Abstract:We propose an optimized land subsidence prediction method combining empirical wavelet Transform(EWT) and Prophet prediction model to solve the problems of low prediction accuracy of single traditional method and unstable prediction process. Taking the Dexing mining area in Shangrao City, Jiangxi Province, as an example, using the 30-view sentinel No.1 image, we carry out the SBAS-InSAR subsidence study and obtain the time series data of the subsidence of the area in the study time span. This method first performs EWT adaptive decomposition on the original sedimentation time series data, decomposes to produce empirical scale components and a series of empirical wavelet components, performs Prophet prediction and superimposition reconstruction on each component to obtain the final predicted settlement data, and finally adopts retrospective prediction to verify the prediction accuracy and reliability of the proposed method. Experimental results show that the EWT-Prophet combined model is overall better than the single Prophet model and the traditional ARMA model. Compared with the other two methods, the EWT-Prophet model improves the root mean square error by 51.53% and 59.03% respectively and the average percentage error is increased by 57.81% and 64.85% respectively, indicating that the prediction effect of this method is better, has better applicability, and provides an effective method for the prediction of large-area mining area settlement.