According to the small sample and nonlinear characteristics of the slope measurement data of foundation pit, a rolling prediction method of time series based on GA-LSSVM model for foundation pit wall measurement was presented.The cubic spline interpolation method was used to pretreat the time series of the foundation pit.Genetic algorithm (GA) was used to optimize the parameters in least squares support vector machine (LSSVM).The optimal parameter model was found out, and the GA-LSSVM time series rolling prediction model was established.The prediction results were evaluated by correlation coefficient R and the mean square error (Mean Squared Error, MSE).The method was applied to the prediction and analysis of the pit excavation of a subway station in Guangzhou, and it was compared with the least squares support vector machine (LSSVM) model without parameter optimization.The results showed that the correlation coefficient of the prediction model was high, the mean square error was small, the prediction result was more accurate.It is of great significance to improve the safety of the construction of the foundation pit and similar projects.