Abstract: To improve the accuracy of predicting the displacement of Baishuihe landslide, we present a prediction model combining the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), bat algorithm (BA), support vector regression machine (SVR), and Adaptive Boosting (Adaboost). By using CEEMDAN, we decompose the displacement of Baishuihe landslide into trend term and fluctuation term composed of IMF sub-items. BP neural network is used to predict the trend term displacement, and the CEEMDAN-BA-SVR-Adaboost model to predict the fluctuation term. We further compare the present prediction result with those of CEEMDAN-PSO-SVR-Adaboost, CEEMDAN-BA-BP-Adaboost, CEEMADAN-BA-SVR, as well as BA-SVR-Adaboost models to verify the superiority of the present method in displacement prediction. We employ the CEEMDAN-BA-SVR-Adaboost model to predict the fluctuation term displacement of monitoring point ZG118 while calculating the final cumulative predicted displacement of monitoring point ZG93. Results verify that the CEEMDAN-BA-SVR-Adaboost model has good accuracy and applicability in predicting the displacement of Baishuihe landslide.

Key words: Baishuihe landslide, displacement prediction, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), bat algorithm (BA), support vector regression machine (SVR), Adaboost