Abstract: It is of great significance to analyze the evolution characteristics of landslide cumulative displacement for displacement prediction. According to the displacement curve of different growth trends, we developed a suitable model to predict the displacement, which will effectively improve the accuracy of the prediction result. In this paper, by analyzing the trend of cumulative displacement of step landslide in the Three Gorges Reservoir area, the displacement curve of step landslide can be divided into four types: equal amplitude-step type, decrease amplitude-step type, increase-step type and compound type, as well as we established the combined prediction model of time series. Taking the monitoring point ZG111 of Bazhimen landslide and ZG326 of Baijiabao landslide as examples, in the light of the principle of time series, the cumulative displacement is decomposed into trend displacement, periodic displacement and random displacement by variational mode decomposition(VMD)method. The trend displacement was modeled and analyzed by the methods of linear regression with one variable and nonlinear regression with power function, and the prediction results were predicted by the weighted least square method(WLS). The Sparrow search algorithm(SSA) is used to optimize the BP neural network model, and combined with the idea of rolling prediction, so we can predict the periodic displacement and random displacement. The cumulative displacement prediction results are the sum of the predicted values. The results show that the MAPE of the trend displacement prediction is 1.2% and 0.77%, respectively. The fitting effect of periodic displacement and random displacement is good, and the prediction results are in line with the overall trend of displacement. The predicted MAPE of cumulative displacement is within 2%, and the predicted results are in good agreement with the actual values. The prediction model presented in this paper meets the requirement of prediction accuracy, realizes the prediction of landslide displacement in the future, has strong practical value in engineering, and provides guidance for the research work of landslide disaster prediction and prevention.