Abstract: Ground settlement induced during deep foundation pit excavation has a significant impact on the overall stability of the foundation pit project and the surrounding environment. This paper takes a deep foundation pit project in Yan'an City as a case study, using the BP(back propagation) artificial neural network model to learn and predict ground settlement data. Pridicted results are compared with on-site monitoring data to verify their reliability, and the influence of various feature factors on the prediction model is analysed. The results indicate that early settlement is significant and occurs at a rapid rate, jointly controlling the overall settlement of the excavation pit with soil rebound effects. The BP model can effectively predict ground settlement during deep foundation pit excavation and the predicted results for each monitoring point are closely aligned with actual measurements, demonstrating high accuracy.Excavation depth and anchor cable tension are key factors influencing settlement prediction, having higher weights and sensitivity during model training, so special attention should be paid to them during actual prediction. The research findings of this study can provide theoretical support for settlement pridiction and construction control in similar deep foundation pit engineering.