There is a complex nonlinear relationship between landslide influencing factors and landslide displacement, and predicting landslide displacement is helpful to study this nonlinear relationship. In the Three Gorges Reservoir Area, many landslides are affected by seasonal rainfall and periodic fluctuation of reservoir water level, and the displacement curve shows step-wise increase. Therefore, the landslide displacement can be decomposed into linearly increasing trend terms and fluctuating periodic terms. Bazimen landslide is a typical step-wise landslide in the Three Gorges Reservoir area, which is used to compare the accuracy of three machine learning methods. A polynomial function was used to predict trend displacement, and periodic displacement was predicted by using three machine learning models which were ASF-SVR, random forests, and gradient lifting trees. The RMSE of the three models is 28.11mm, 20.41mm, and 22.95mm respectively. The results show that the prediction effect of random forest is the best.