In order to effectively monitor the stability data of foundation pit during the construction process, the study proposes a monitoring model that can monitor the deformation state of foundation pit in subway stations in real time. The study analyzes the force state of the support through the deformation mode and extrusion direction of the soil layer, and calculates the displacement and deformation of the soil layer by combining with the hydrological environment around the soil layer. Data monitoring points are set up to monitor the dynamic data of key stress areas. The study was conducted to predict the deformation state of foundation pit with long short-term memory artificial neural network model, combined with fast Lagrangian analysis of continuum for numerical validation. The study revealed that the accuracy of prediction of test data with the research monitoring model was 0.912. The convergence speed of the research model during training was 37.34%-38.27% higher than other methods. Meanwhile, the data occupancy of the research model in the prediction process was low and its redundancy was reduced by 10.74%-36.17% compared to other methods. During the data prediction process of foundation pit, the mean accuracy of the research method for the prediction of wall deformation and settlement of foundation pit remained above 95%. Therefore, the research method provides an efficient monitoring model for the safety management of foundation pit design in subway stations.