Based on SBAS-InSAR technology, 47 sentinel-1A images from January 2018 to December 2021 were used to monitor the subsidence of Yaojie mining area. According to the subsidence characteristics of mining areas, 4 prediction models, GM(2,1), BP neural network, PSO-SVR and LSTM, are constructed to predict and analyze the subsidence values of deformation centers in three mining areas from 2018 to 2021. The results show that most areas of Yaojie Mining Area are in a stable state, and the obvious subsidence funnel is formed in the center of the mining area. The maximum annual average subsidence rate is 157.01 mm/a, and the maximum accumulated subsidence from 2020 to 2021 is 681.82 mm, so there is a danger of large-scale subsidence and secondary geological disasters in mined-out areas. The prediction results show that the four models show different prediction accuracy, and the LSTM model has the highest prediction accuracy, which can be used as a relatively reliable prediction model of land subsidence in mining areas.