Application of time series analysis and PSO–SVM model in predicting the Bazimen landslide in the Three Gorges Reservoir, China

•The deformation and failure mechanism of the Bazimen landslide are analyzed.•The relationship of landslide periodic displacement and the controlling factors of landslide deformation are analyzed.•Particle Swarm Optimization and Support Vector Machine are integrated and applied in landslide displacement prediction.•A landslide displacement prediction model is proposed in relation to controlling factors.

The landslide displacement in the Three Gorges Reservoir, China, experiences step-like deformation that is influenced by rainfall and the periodic scheduling of the reservoir. In view of the step-like characteristic, the Particle Swarm Optimization and Support Vector Machine (PSO–SVM) coupling model based on the response of the induced factors was proposed to predict the landslide displacement. The moving average method was adopted to divide the total displacement into trend term and periodic term. The trend displacement was controlled by the geological conditions and predicted by polynomial function, while the periodic displacement was under the combined control of the triggers and the evolution state of the landslide. Therefore, the PSO–SVM model, based on the factors of the precipitation, the variation range of the reservoir and the displacements of the prior-periods, was proposed to predict the periodic displacement. The typical step-like landslide in the Three Gorges Reservoir, which is known as the Bazimen landslide, was taken as a case study to verify the prediction results. The values of the root mean square error and the mean absolute percentage error were 13.28 and 25.95, respectively. The results showed that rainfall and reservoir water level were the dominant factors for the step-like landslide deformation. The evolution state of the landslide was also significant in reflecting the response relationship between the displacement and inducing factors. In conclusion, the proposed PSO–SVM model can better represent the response relationship between the factors and the periodic displacement, which made the predicted values of the total displacement fit with the measured values greatly.