Abstract: Landslide hazards occur frequently in China, but the prediction and forecast of landslide deformation have always been a problem. Therefore, the deformation and failure of landslides caused major property damage and loss of human lives each year. Based on the monitoring data for more than ten years, we studied and analyzed the deformation characteristics of the Bazimen landslide. The main reasons of the deformation are the decline of reservoir water level and the rainfall. Moreover the cumulative displacement curve has the "step type" deformation characteristic. When we remove or reduce the external factors, the cumulative displacement-time curve becomes smooth. According to this characteristic, this paper selects the monitoring data of the annual step of the deformation curve(June-August), and uses the cumulative displacement as the objective function. Based on the one-variable linear regression model, the landslide monitoring data of the Bazimen landslide from 2004 to 2017 are analyzed. The results show that the one-variable linear regression model can simulate the deformation process of the "step segment" of the Bazimen landslide well. The cumulative displacement with time in this deformation phase is linear, and the slope of the line is basically the same. According to this linear relationship, we can predict the cumulative displacement of the landslide. The results show that prediction error is less than ±5 mm comparing with the actual monitoring data, and the relative error is below 1%. The accuracy can meet the requirements of landslide monitoring and early warning. It can be used for the prevention of the landslide work and provides reference.