Due to the normal forecasting methods for subgrade settlement using observation data have differentapplications，and the predicting results has bigger volatility and lower accuracy. The Combined forecasting model ofsubgrade settlement based on Least Square Twin Support Vector Ｒegression ( LSTSVＲ) is proposed in this paper. Itscore is that the growth curves with the S-type characteristics are treated as single forecasting model according to thebasic settlement law of subgrade and characteristics of settlement curve. Considering prediction results of eachindividual model as the least square support vector regression model input and the combined forecasting model ofsubgrade settlement is constructed. The result of engineering practice shows that the proposed method has betterprediction accuracy and stability