Aiming at the main safety problems of shield tunnel construction of Guangzhou Metro Line 18, which are presented by the surface subsidence and segment floating, the deep learning network-deep belief network is used to construct the prediction model for the construction safety based on the big data between construction parameters and safety problems obtained from field monitoring. The influence of six main parameters on construction safety is studied. The results show that the deep belief network can obtain more accurate prediction values of surface settlement and segment floating based on the big data of field monitoring. Moreover, the surface settlement decreases with the increase of the soil bin pressure, grouting volume and grouting pressure, and increases with the increase of the jack thrust, cutter torque and tunneling speed. The floating capacity of the segment increases with the increase of the soil bin pressure and grouting pressure, and decreases with the increase of the jack thrust and grouting volume. The torque of the cutter head and the tunneling speed have little effect on the floating of the segment.