LANDSLIDE DISPLACEMENT PREDICTION BASED ON MULTIPLE DATA-DRIVEN MODEL METHODS

Slope displacement is a macroscopic manifestation of landslide evolution. Analyzing and predicting landslide displacement are of great significance for disaster prevention and mitigation. Since the landslide displacement is obvious nonlinear characteristics, single model is often difficult to delineate the complexity and nonlinearity of landslide displacement. To find a universal method for predicting landslide displacement, a new method combined with multiple data-driven modeling methods is proposed to predict the landslide displacement. The new method is based on time series analysis. The landslide displacement sequence is decomposed into trend term and periodic term. The trend term is treated using parallel grey neural network, and uses the artificial bee conony(ABC)to find the optimal extreme learning machine model(ELM)to predict the periodic term. This paper takes Baishuihe landslide and Bazimen landslide for examples. After statistically analyzing the displacement data, the gray neural network model predicts trending displacement, and the optimized learning machine model train and predict the periodic term. The result shows that the optimized extreme learning machine model is better than the extreme learning machine model and wavelet neural network. Therefore, the proposed combination of grey neural network and ABC-ELM can be used as a reference for practical engineering.