PSO-Prophet model for slope deformation analysis and prediction
WANG Zhiying1,LI Zongchun1,XU Wenxue2
(1. Institute of Geospatial Information,PLA Strategic Support Force Information Engineering University,Zhengzhou,Henan 450001,China;2. Institute of Engineering Design,Air Force Academy,Beijing 100077,China)
Abstract:The PSO-Prophet model for slope deformation analysis and prediction is constructed to solve the problems of highly uncertainty of periodic term extraction and the increasing complexity of combination forecasting model. The Prophet model,with the function of time series decomposition and combination prediction,is introduced to deal with the slope deformation signal. What?s more,the PSO algorithm is used to optimized the parameters to improve the adaptability and prediction accuracy of the model. The prediction experiments of slope monitoring points are carried out,and the prediction accuracy of PSO-Prophet model is 46% higher than Prophet model in the original unequal interval time series. In the combination prediction of equal interval time series after interpolation,the root-mean-square errors of the PSO-Prophet model,the S-GSSVR model,the HP-ARIMA-LSTM model and SSA-Verhulst-FS model are 4.7,9.6,13.2 and 12.0 mm respectively. The experimental results show that the PSO-Prophet model has a simple process and higher prediction accuracy for slope deformation time series greatly affected by periodic term.