<p>The monitoring system plays a crucial role in the stability analysis of ground subsidence caused by mining activities and in predicting associated engineering geological disasters. The dynamic evolution characteristics of ground subsidence and the presence of uncertain noise in monitoring data pose considerable challenges to the analysis of its patterns and the real-time prediction and early warning of such events. A ground subsidence optimization model was proposed based on GPS monitoring data of the Hainan-Shilu Iron Mine in China, a mine transitioning from open-pit to underground mining. The proposed model combined wavelet threshold filtering (WTF), Kalman filtering technique (KFT), and Fast Fourier Transform (FFT), and was referred to as the WTF-KFT-FFT model. First, the ground subsidence sequence was decomposed into a trend sequence and a fluctuation sequence by WTF. The functional expression of the trend sequence was fitted with a quadratic polynomial. A dynamic process noise KFT was proposed to filter noise from the fluctuation sequence. The functional expression of the filtered fluctuation sequence was derived using FFT. Finally, the optimization and prediction of the ground subsidence sequence were achieved by superimposing the functional expressions of the trend and fluctuation sequences. The proposed optimization model was validated by comparing it with existing monitoring data and conventional optimization methods. The results demonstrated good optimization and prediction accuracy for the ground subsidence sequence at the Hainan-Shilu Iron Mine. Based on the computational results, the subsidence patterns were carefully analyzed, and suggestions for improving the monitoring scheme were made, which provided valuable support for the slope early warning system. These efforts also contributed to improving both the safety and sustainability of mining resource extraction.</p>