<p>To ensure construction safety during tunnel excavation, it is critical to accurately and efficiently determine stratigraphic parameters and predict surface settlement characteristics. Currently, the back analysis method is widely used as an indirect method to determine stratigraphic parameters. However, most back analysis methods require a large number of numerical calculation models, resulting in a large computational cost. To improve the accuracy and efficiency of stratigraphic parameter back analysis, this paper presents the beluga whale optimization-least squares support vector regression-FLAC^3D (BWO-LSSVR-FLAC^3D) method. In this method, the error between the simulated settlement and actual monitored settlement is the objective function, and the stratigraphic parameters are the optimization variables. The BWO algorithm searches for the global minimum of the objective function, and the LSSVR algorithm replaces the time-consuming numerical calculations to efficiently and accurately back analyze the corresponding stratigraphic parameters. In addition, the LSSVR algorithm is used to train a surrogate model that can predict the surface settlement caused by the tunnel excavation process. The results show that compared with the BWO-FLAC^3D and the particle swarm optimization-support vector regression-FLAC^3D (PSO-SVR-FLAC^3D) algorithms, the BWO-LSSVR-FLAC^3D method greatly improves calculation efficiency while ensuring accuracy. Moreover, the prediction model trained by the LSSVR algorithm performs better. Compared with the random forest and support vector regression surrogate models, it has higher prediction accuracy in engineering applications. Note that the proposed method can be used as a complement to rather than a replacement for existing stratigraphic parameters back analysis and surface settlement prediction in tunnel construction.</p>