A hybrid predicting model for displacement of multifactor-triggered landslides

This paper presents a new hybrid model for land-slide distance prediction. In the model, the cumulative displacement are divided into three parts: the trend term, the period term, and the random noise obtained by the wavelet domain de-nosing method and Hodrick-Prescott (HP) filter. The trend term controlled by the geological conditions is generated using the double exponential smoothing (DES). The period term is predicted by the extreme learning machine (ELM) model, and the dynamic multi-swarm particle swarm optimizer (DMS-PSO) algorithm is applied to obtain optimal parameters of ELM. Case study involving real data collected from the Baishuihe landslide in China is used to verify that the hybrid approach enhances the ability to calculate the period term. Inputs of the proposed model include the period factors extracted from the seasonal triggers and displacement values which enhance excellently the robustness of the prediction model of the period displacement. Extensive experiments are carried out on the Baishuihe landslide dates. Comparing with the predictions obtained by the real original displacement, our model is efficient for predicting the landslide distance of multiple factors induced landslide.