Tunneling-induced ground surface settlement prediction based on relevance vector machine

Ground surface settlement prediction is very important to identify potential damage incurred to adjacent structures. However, the traditional prediction methods, such as empirical formulae, are inaccurate because most of them do not take into consideration all the relevant factors. In this paper, the relevance vector machine (RVM) is introduced to predict the unseen data. Thus we focus on the use of RVM for regression. Theoretically, tunneling-induced ground surface settlement can be regarded as a time sequence. Therefore different data series are sampled from different sensors and is taken as the training set for RVM to obtain the non-linear model. The data prediction of new sensors based on the established model showed that the model is effective and applicable.