Abstract: We analyze the accuracy trends in landslide deformation, based on the exponential smoothing method and the practical stage of evolution landslide. We establish a connection between the main predisposing factor and model parameters, introducing monthly cumulative rainfall as the evaluation factor for the dynamic model parameter. We use cumulative displacement data from Baijiabao landslide for fitting and forecasting. The result shows that the absolute error and correlation coefficients in our final modal were 11.346 and 0.933. Compared with the conventional method using static parameters, the proposed more in line with the general laws of development of rainfall-induced landslides, and is more accurate.