Improved Support Vector Machine Regression in Multi-Step-Ahead Prediction for Tunnel ‎Surrounding Rock Displacement ‎

A dependable long-term prediction of tunnel surrounding rock displacement is an effective way to predict the rock displacement values into the future. A multi-step-ahead prediction model, which is based on support vector machine (SVM), is proposed for tunnel surrounding rock displacement prediction. To improve the performance of SVM, parameter identification is used for SVM. In addition, to treat with the time-varying features of tunnel surrounding rock displacement, a forgetting factor is introduced to adjust the weights between new and old data.At last, the data from the Chijiangchong tunnel are selected to examine the performance of the prediction model. Comparative results were presented between SVMFF (SVM with a forgetting factor) and artificial neural network with a forgetting factor (ANNFF) show that SVMFF is generally better than ANNFF. This indicates that a forgetting factor can effectively improve the performance of SVM, especially for the time-varying problems.