Aiming at the shortcomings of the current building settlement prediction model triggered by pit excavation, such as insufficient accuracy, slow convergence speed, and easy to fall into local optimization, a building settlement prediction model based on the improved aquila optimizer (IAO) optimized least squares support vector machine (LSSVM) was proposed. Tent chaotic mapping was utilized to improve the population diversity level of the aquila optimizer, and then adaptive weighting was used to strengthen the algorithm's full-stage optimization search capability. The IAO algorithm was introduced to optimize the regularization parameter and kernel function width of LSSVM to construct a building settlement prediction model based on IAO-LSSVM, and the prediction model was verified in a subway pit project in Huaqiang South, Shenzhen. The results show that the settlement prediction model has higher accuracy, faster convergence and stronger ability to jump out of the local optimization domain than the traditional prediction model. The predicted values of the model are in good agreement with the actual settlement monitoring values, and its error is around 5 %, which is more suitable for predicting the settlement of the surrounding buildings caused by the excavation of the subway foundation pit in the city.