A wavelet intelligence prediction system (WIPS) is presented herein to predict the ground deformations induced by tunneling. In this method, the solution is comprised of three parts: wavelet analysis, model identification and system prediction. Based on the sensitivity analysis of influencing factors, ground deformation is decomposed into the trend deformation and the wave deformation. Wavelet analysis is introduced to filter the residual error and extract the actual deformations, which is similar to de-noising in signal processing. In addition, the identification model is established by using Elman neural network based on modified PSO (named EMPIM), with which one can approximate the actual deformations. The prediction system (i.e.; WIPS) developed with two identifiers enable one to map all influencing parameters to ground deformations, which helps avoid complex theoretical analysis of rock-soil mechanisms and mathematical descriptions of ground deformations. Later, WIPS is applied to estimate future deformations. The validation use cases show that the WIPS is an effective tool for predicting ground deformations dynamically under difficult and uncertain conditions, and can be widely applied to practical subway projects.