Abstract: Aiming at the deficiency of accuracy,stability and generalization of conventional concrete dam safety monitoring models,combined with the advantages of neighborhood rough set(NRS)theory in data attribute reduction and redundant information elimination,and advantages of random forest(RF)method in analyzing nonlinear,highly collinearity and noisy data,a concrete dam deformation monitoring model based on NRS-RF was built.In this paper,taking the Zhouning Hydropower Station as a case,15 initial influencing factors were reduced to 5 core influencing factors by NRS,and the reduced attributes were used as input variables of RF to establish a concrete dam deformation monitoring model.The prediction results of this model were compared with the that of model based on the traditional least square method(OLS).The results showed that the fitting and prediction accuracy of the concrete dam deformation monitoring model based on NRS-RF was high with good stability.The study provides a new method for dam deformation monitoring.