Verhulst Inverse-Function Residual Correction Model for Landslide Prediction and Forecast
Liu Bao-chen · Zhongguo tiedao kexue · 2009
Verhulst inverse-function model is suitable for dealing with such prediction problems as non-negative,monotony and with small fluctuation of the monitoring data.Therefore it is poor for fitting the data sequence that fluctuates at random and is low in forecast accuracy.BP neural network prediction model is applicable to the random,nonlinear and dynamic data system,which can make up the limitation of Verhulst inverse-function prediction and forecast model.On the basis of Verhulst inverse-function prediction and forecast model,and in combination with BP neural network residual series,a new residual correction model of Verhulst inverse-function is built for landslide prediction and forecast.The prediction method of Verhulst inverse-function residual correction model for predicting the time of landslide instability is studied.Both prediction methods of Verhulst inverse-function residual correction model and Verhulst inverse-function model are used to predict the actual landslide cases home and abroad.The research indicates that the prediction and forecasting results of Verhulst inverse-function residual correction model are closer to the practical observation data.The average relative error of the prediction values obtained by Verhulst inverse-function residual correction model is reduced 1%～7% compared with that obtained by Verhulst inverse-function model.It shows that Verhulst inverse-function residual correction model has a higher precision and more extensive application for landslide prediction and forecast.