Time Sequence Forecast of Ground Settlement Based on WT-SVR-ARMA

Pengfei Wang

Abstract: Ground settlement caused by the subway underground excavation is a very critical issue. Considering the important significance of an accurate ground settlement prediction model to underground construction, a time series prediction method based on wavelet transform (WT) and support vector regression (SVR) and autoregressive moving average (ARMA) model to predict ground settlement is prosed. Wavelet transform is used to decompose and reconstruct ground settlement time sequence into trend sequence and random sequence. Trend sequence is forecasted by SVR model. Random sequence is forecasted by ARMA model. The final prediction values are the sum of trend sequence and random sequence prediction values. This method is used to the data sampled from sensors located at Ziyou Road station in Changchun and it shows that this method is valid and applicable.

Keywords: time sequence, wavelet transform, support vector regression, autoregressive moving average, ground settlement