To solve the engineering problem of unclear standards and strong subjective experience when shield tunneling drivers set excavation parameters, which makes it difficult to control the shield tunneling attitude, an intelligent prediction model for shield tunneling attitude that considers the comprehensive effect of geological conditions, tunnel structure, and excavation parameters was proposed. Firstly, AWPSO (adaptive inertia weight particle swarm optimization) algorithm was established. Then, a shield attitude prediction model was constructed by combining GRU (gated recurrent unit) neural network, where the AWPSO algorithm was used to determine the optimal combination of hyperparameters in the GRU neural network. Finally, a case study was conducted to verify the on-site monitoring data between Zhangjiang Station and Resort Station on the Shanghai Suburban Railway Airport Connection Line. The results indicate that the proposed shield tunneling attitude prediction model based on AWPSO-GRU has high reliability and engineering practicality, which can provide reference and basis for setting construction parameters during shield tunneling.