Large-diameter shield tunneling under complex stratum conditions faces the challenge of multi-objective collaborative optimization of tunneling efficiency, energy consumption, and surface settlement. A multi-objective optimization framework based on data augmentation was proposed. The framework innovatively integrated the natural gradient boosting (NGBoost) probabilistic prediction model, the non-dominated sorting genetic algorithm III (NSGA-III) evolutionary algorithm, and the technique for order preference by similarity to an ideal solution (TOPSIS) method. The TOPSIS method was used to select the optimal combination of shield operating parameters from the Pareto optimal solution set. An actual shield tunneling project in Hangzhou was taken as a case study. Research on the optimization of shield operating parameters was carried out under scenarios with different burial depths and soil mechanical properties. Experimental results demonstrate that high prediction accuracy for tunneling speed, specific energy consumption, and surface settlement is achieved by the proposed method: the root mean square errors (RMSE) between predicted and actual results are 3.4348 mm/min, 9.1662 MJ/m3, and 5.8210 mm, respectively, while the coefficient of determination (R2) reaching 0.7759, 0.8444, and 0.9026. The parameter optimization effect is significant: compared with the pre-optimization state, the tunneling speed is increased by 19.81%, the specific energy consumption is reduced by 62.09%, and the surface settlement is decreased by 21.53%. In conclusion, the proposed multi-objective optimization framework can effectively balance efficiency, energy consumption, and safety indicators in large-diameter shield tunneling. Under complex stratum conditions, it significantly improves construction efficiency and safety level, and reduces construction energy consumption. This framework provides a reliable technical reference for parameter optimization in similar engineering projects.