Deformation monitoring data can reflect the service behavior of ultra-high arch dams directly, they contain abundant spatio-temporal information and evolution law, which is of great significance to the long-term stability of the project. However, the multi-source and multi-dimensional deformation monitoring data are often affected by the monitoring instrument itself and other external factors, which will cause interference to the following data analysis. Regarding to the missing data in dam deformation monitoring sequence, the target monitoring sites with strong correlation were obtained based on the spatial correlation of monitoring site deformation calculated by Apriori association rule, and then the deformation monitoring data were adopted as the input samples to fill in the gaps in deformation monitoring sequence of the target monitoring sites using Bayesian optimized XGBoost regression model. The case study of Jinping ultra-high arch dam shows that this method can fill in the deformation monitoring gaps efficiently and accurately, so it is applicable to the filling of missing deformation data of similar dam projects.