English：

Subway construction projects often face the risk of overruns of cost and delays of constructionperiod due to complex geological conditions and dynamic parameter coupling. In this study, a model forcost and construction period prediction was constructed based on an artificial neural network （ANN），and the back propagation （ANN⁃BP） algorithm was used to optimize the network structure and activationfunction. Through the verification of engineering data of Jinan Metro, the research results indicate thatthe model of hyperbolic tangent function, combined with 5 neuron hidden layers, shows excellentperformance in cost and construction period prediction. Specifically, the coefficient of determination（R2） of the test set for cost prediction is 0.899, and the root mean square error （RMSE） is 0.028. Thetest set R2of the construction period prediction is 0.971, and the RMSE is 0.024,which is significantlyreduced by 53. 8% compared with the Sigmoid function. Compared with support vector regression （SVR）and random forest （RF） models, R2of ANN⁃BP in cost prediction is increased by 18.7% and 9.2%,respectively, and RMSE is decreased by 32.4% and 22.2%, respectively. In the construction periodprediction, R2is increased by 21.5% and 14.0%, respectively, and RMSE is decreased by 28. 9% and22. 6%, respectively. The model effectively captures the nonlinear relationship between surrounding rockgrade and thrust torque by integrating geotechnical mechanical parameters and shield dynamic workingcondition data, and the error is concentrated in the range of ± 5%. This model provides a high⁃precision tool for cost control and schedule management of tunnel engineering, and its algorithm robustness hassignificant application value in complex geological scenarios.