Landslide Displacement Prediction of WA-SVM Coupling Model Based on Chaotic Sequence

Confronted with the chaotic characteristics of landslide displacement and the deficiencies of traditional time series prediction model, the wavelet analysis -support vector machine model (WA-SVM) based on chaotic time series for landslide displacement prediction is proposed. On the basis of the analysis of chaotic characteristics, landslide displacement is decomposed in to components with different frequency characteristics by wavelet analysis; Then each component is phase space reconstructed and predicted by the support vector machine (SVM) separately; Finally, the predicted value of the original sequence is obtained by superimposing all the components. Bazimen landslide in the Three Gorges Reservoir area is taken as an example, and a comparison between the proposed model and the wavelet analysisBP Neural Network (WA-BP) and single SVM model is implemented, where the root mean square error (RMSE) of WA-SVM model is 10.45, the mean absolute percentage error (MAPE) is 1.06%, and the correlation coefficient is 0.989, which are better than the WA-BP and single SVM model. Results show that, the WA-SVM coupling model based on chaotic sequence is of high prediction accuracy, which is an effective prediction model for landslide displacement.