Abstract: For the common mode error in GPS deformation monitoring coordinate sequence of mining subsidence, the principal component analysis (PCA) was used to extract and eliminate the correlative error among stations in high frequent single epoch algorithm. Welch spectrum method is used to analyze the cycle frequency characteristics of using PCA space filtering to eliminate common-mode error. Finally, we process real-time settlement observation data from GPS deformation monitoring network in a certain mining area, and verify the feasibility of removing common-mode error by PCA spatial filtering method. The method improves the precision of mining subsidence with GPS observation.