Abstract: Aiming at the nonlinear and random characteristics of building settlement,a Richards curve prediction model based on Markov chain correction was established to improve the prediction accuracy of the model.First,the fminsearch function of Matlab platform was used to obtain the optimal parameter solution of Richards model,based on the principle of minimizing the sum of squared error between the fitted value and the measured value.Then,the residual error was corrected by Markov's method,finally the combination model was used to predict the settlement.The prediction accuracy of the model was compared and analyzed by examples.The results show that it was effective to estimate the Richards parameters by using fminsearch.The average absolute percentage error and the root mean square error of the Richards model optimized by Markov chain were smaller than the single model prediction,and the prediction accuracy was improved.The combined model prediction results can be better reflect the overall trends of data series development and changes over time as well as the inherent laws between the various states of the system,and it is suitable for building settlement prediction.

Key words: Markov chain, Richards curve, settlement prediction, fminserch