Abstract: Because the time series data of dam displacement is affected by various complex factors and has non-stationary and nonlinear characteristics,it is difficult to accurately describe the complex laws of dam displacement and deformation using a single traditional time series prediction model. Considering of the comprehensive consideration of the nonlinear and linear characteristics of the dam displacement time series,this paper proposes a time series prediction model based on the combination of SVM and ARIMA. The dam deformation time series is divided into the nonlinear part and linear part. For the nonlinear part,the support vector machine is used for rolling prediction. Compared with the NAR dynamic neural network,the support vector machine has a relative advantage in dealing with nonlinear problems. For the linear part,a single-step rolling prediction is made through the ARIMA model,and the prediction value of the combined model is obtained by combining the two prediction results. The combined model is tested with the measured data of the dam. The experimental results show that the SVM-ARIMA combined model has high prediction accuracy and can better describe the change trend of the dam displacement, which has certain practical value.

Key words: dam deformation, SVM, ARIMA model, SVM-ARIMA combined model, rolling prediction