Study on Combined Prediction Method of Post-construction Settlement of Loess High Fill Based on Inclusion Test
Aiming at the problems of single model selection and weight coefficient distribution when es-tablishing the combined prediction model of post-construction settlement of loess high fill site,a com-bined prediction method based on inclusiveness test and optimal weighting method was proposed.Based on the measured data of post-construction settlement,seven single prediction models such as hyperbolic function and logarithmic function were established.According to the prediction accuracy of each single model,the order was sorted,and the inclusion test was carried out step by step to select the appropriate single model.Different weight coefficient distribution methods were used to combine the selected models,and the best combination method was selected based on the principle of minimi-zing the inner fitting error and extrapolation prediction error.The results show that the suitable single model can be screened out by the idea of step-by-step inclusion test according to the order of advanta-ges and disadvantages,and the best combination prediction effect of single model is achieved by using the optimal weighting method.Compared with the traditional forecasting method,it reduces the num-ber of single models participating in the combination and improves the forecasting accuracy and effi-ciency.
万方数据
维普
