Abstract: This paper aims to address the deficiencies in surface subsidence monitoring accuracy and the limited adaptability of time-series prediction models in open-pit mines. To this end, a novel subsidence analysis and prediction method for open-pit mines is proposed. This method integrates time-series InSAR technology with an enhanced long- and short-term memory（LSTM） model. Firstly, SBAS-InSAR technology is employed to process 59-view Sentinel-1 satellite images, thereby acquiring the spatial distribution characteristics of the overall deformation in the mining area and the annual average subsidence rate at millimeter level. Secondly, the traditional LSTM model is optimized and improved, and encoder and decoder architectures are introduced to construct the framework of subsidence prediction. The results demonstrate that: the SBAS-InSAR monitoring accuracy is high; the prediction accuracy of the enhanced LSTM model is notably enhanced, with an average absolute error of 1.946 mm and an average root-mean-square error of 2.453 mm for the four points; the average absolute error of 3.670 mm and the average root-mean-square error of 4.560 mm for the four points of the traditional LSTM model. In comparison with the traditional LSTM model, the average absolute error and average root mean square error are reduced by at least 46.98% and 46.21%, respectively. Consequently, the integration of time-series InSAR and enhanced LSTM demonstrates a substantial application efficacy in the analysis and prediction of open-pit mine subsidence.