Abstract：
BP neural network and adaptive filter combined model was built with variance theory according to the characteristic of simulation and forecast value from independent BP neural network model and adaptive filter model fluctuating relatively to the true value. The combined model was used to analyze and forecast the land subsidence around foundation of the third period project of Tianjin Fuyu Square. High accuracy of the combined model was proved by comparing the combined model data with the results from independent BP neural network model and adaptive filter model.
