In order to ensure flight safety, the requirement of foundation settlement of high-plateau airport is stricter than that of airport in plain area. In order to monitor the abnormal state of runway foundation in the process of use of a high-plateau airport and prevent and resolve the major risk of foundation settlement to flight safety. It takes a high-plateau airport in the southwest mountain area as an example, selecting two representative A and B sections for analysis. It takes the first 60 days’ monitoring data as training samples, which shows nonlinear characteristics. The Long and Short Term Memory neural network (LSTM) prediction model and BP neural network model are constructed to predict the trend of foundation settlement after construction. In the process of building the LSTM model, the minimum root-mean-square error of test samples was selected as the fitness function, and the parameters of the LSTM model were modified by Genetic Algorithm (GA). And then the modified LSTM prediction model based on the early settlement of the foundation was constructed. The results shows that the modified LSTM model and BP model constructed in this paper are generally consistent with the field measured values in the prediction of airport foundation settlement of high-plateau, but the modified LSTM model is more sensitive to the abrupt change of data and has a more stable trend than the BP model. The predicted values of the modified LSTM model are all greater than those of the BP model, and the predicted values of the modified LSTM model are closer to the monitored values in the field than the predicted values of the BP model, and the relative error between the predicted values and the monitored values is less than 3%. The research can provide a reliable theoretical reference for the design, construction, operation management and later maintenance of high-plateau airport.