When tunneling with a tunnel boring machine (TBM) in permeable soil, excess pore-water pressure are inevitably generated in the soil surrounding the TBM. Because excess pore-water pressure reduce the effective face support pressure, accurately predicting their magnitude is important for determining the required effective face support pressure. In this study, a long - - short-term memory (LSTM)-based deep learning model is employed to predict variations in pore-water pressure generated by TBM tunneling using time-series data derived from field monitoring data and TBM data collected during construction of the Green Hart Tunnel (GHT) in the Netherlands. Four obtainable input variables are selected to quantify pore-water pressure at two monitoring points that have different distances (8.3 and 107 m) along the transverse axis. Three accuracy metrics are introduced to evaluate the performance of two prediction tasks, with input variables' importance on the output ranked according to their corresponding sensitivity values. It demonstrates that the proposed LSTM-based deep learning model can accurately predict the pore-water pressure ahead of the TBM in drilling-standstill cycles, which can further serve as a tool for TBM operators to use in assessing real-time tunnel face stability.