针对滑坡位移序列的混沌特性和传统时间序列预测模型的不足，提出了一种基于混沌时间序列的小波分解-极限学习机（WA-ELM）滑坡位移预测模型。该模型以滑坡位移序列混沌特性分析为基础，应用小波分析将位移序列分解为具有不同频率特征的分量，对各特征分量分别进行相空间重构并应用极限学习机进行预测，最后将各特征分量预测值叠加，得到原始位移序列的预测值。以三峡库区八字门滑坡为例，并与小波分析-支持向量机（WA-SVM）以及单独ELM模型进行对比研究。结果表明，基于混沌时间序列的WA-ELM模型预测精度较高且具有较好的通用性与稳定性，是一种有效的滑坡位移预测方法。

To address the chaotic characteristics of landslide displacement sequence and to overcome the deficiency of the traditional time series forecasting models, a WA-ELM prediction model of landslide displacement is proposed based on chaotic time series. The chaotic characteristics of the landslide displacement sequence is analyzed, in which the wavelet analysis(WA) is employed to decompose the displacement sequence into characteristic components with different frequencies. The characteristic components are reconstructed in the phase space and predicted using the extreme learning machine (ELM). Finally, the characteristic components are superposed to obtain the prediction values. A comparative study of Bazimen landslide in Three Gorges Reservoir area is made using WA-SVM and ELM models, respectively. The results show that the predictions of the WA-ELM model based on chaotic time series has higher accuracy and better versatility and stability.