Research on prediction of slope displacement based on a weighted combination forecasting model

Abstract

With the continuous development of slope engineering, people have put forward new requirements for slope safety. Slopes widely exist both above-ground and underground engineering. Landslide prevention is crucial to building and personal safety, and predicting the deformation and evolution trend of landslides is an important part of it. Aiming at the limitation that most of the current landslide displacement prediction models belong to a single model, this paper proposes a weighted combination prediction model to predict the displacement of the slope. Taking the measured displacement of the Three Gorges slope as an example, this paper firstly uses Grey-Marcof, GA-BP neural network and ARIMA model for prediction, and then combines the prediction results of the above three models by the reciprocal variance weighted method, and finally the linear combined prediction model of the slope displacement is established to complete the final prediction. The prediction results show that, compared with the minimum RMSE value of 0.3080 and the maximum RMSE value of 2.1379 predicted by the three methods alone, the combined prediction with RMSE value of 0.1249 has the best prediction accuracy. Moreover, compared with the average relative error of 1.46%, 1.49%, 4.32% and 1.24% of the prediction results by the single method of the authors who provided the data, the prediction results of the combined prediction model also showed a significant improvement in accuracy with the average relative error of 0.51%. Therefore, the weighted combination prediction model can further improve the slope displacement prediction accuracy, and has good scientificity and feasibility.