To improve the prediction accuracy on the soft foundation subsidence model,the paper puts forward L-V model which is superposed on the existing Logistic model and Von-Bertalanffy model,and takes advantage of the Origin fitting software to solve the model parameters.The model is applied to soft foundation subsidence prediction compared with the Logistic model and Von-Bertalanffy model.The results shows that predicted fitting degree of L-V model is the highest to 0.997 8;the L-V model average root error(RMSE=1.512) of settlement later is less than that of Logistic model(RMSE=2.930) and Von-Bertalanffy model(RMSE=3.624),and the prediction is approximated to the observations.