Time Series Prediction Model of Deformation of Foundation Pit Based on Least Squares Support Vector Machines
Xuehong Li · Water Resources and Power · 2011
As traditional neural network suffers from the problems,such as existence of many local minima,the choice of the number of hidden units and overfitting,sliding time window is built and least squares support vector machines(LS-SVM) is proposed to predict the deformation of foundation pit by using measured time series sample.The grid-search method is used to optimize the model parameters and it achieves continuous rolling multi-step prediction of foundation pit deformation.The example results show that the proposed approach outperforms back-propagation(BP) neural network,which has characteristics of good generalization performance and requiring little sample data.