Building settlement forecast using BP neural network

In order to calculate building settlement forecast accurately and construction operation safely, a building settlement forecast method using BP neural network was put forward with deeply analysis of existing building settlement prediction methods. Firstly, BP neural network learning samples are established based on time series analysis method, a three-layer BP neural network is used to settlement forecast, and mean square error and mean absolute percentage error are used to evaluate the precision of the results. Settlement data of Information center building of China University of Mining and Technology (CUMT) is shown as example, the results show that the mean square error of D7 point is 2.5mm; and the mean absolute percentage error is 6.5%; the mean square error of D16 point is 3.4mm; and the mean absolute percentage error is 7%. The forecasting results show that the value predicted by BP neural network conform closely to data measured, the BP neural network model prediction results are accurate, and errors can meet the engineer need. The research provides a new way of Building Settlement Forecast.