This study conducts a comprehensive analysis of the recorded data on ground deformation during the construction of an offshore airport embankment, with the aim of exploring the potential for real-time future prediction and feedback control. Considering the presence of measurement errors and missing data in the observed records, this research investigates the signal reconstruction method based on Dynamic Mode Decomposition (DMD) and examines the feasibility of developing a ground deformation prediction model. The effectiveness of mrDMD, incorporating a multi-resolution representation, is validated for effectively separating local heterogeneity within the observations. Moreover, the introduction of DMD-C (a time-invariant system) highlights its compatibility with linear control theory, thereby demonstrating the feasibility of constructing a future prediction and control model for ground deformation.