A dynamic on-line sliding window support vector machine for tunnel settlement prediction

Aiming at increasing the precision of tunnel settlement prediction, a modified support vector machine (SVM) based on the dynamic on-line sliding window (Dolsw) technique is proposed. In the prediction model, the historically observational settlement data act as the learning samples. The nonlinear relationship between settlement data and influencing variables is established on the basis of on-line learning SVM. In addition, the number of the samples is controlled with dynamic sliding window technique for improving its effectiveness. Finally, the new method can be used to predict the testing samples. Experimental results show that this method can effectively provide reliable predictions with higher precision and greater generalization. Also, it can prevent the over fitting phenomenon.