黄土路基沉降变权重组合模型预测方法研究

Study on Variable-weight Combination Forecasting for Settlement of Loess Subgrade

利用兰武二线路基沉降观测数据,对黄土路基沉降变权重组合模型预测方法进行分析研究。研究结果表明:根据拟合精度选择单个模型时,要尽量选择相对误差偏离方向相反且大小相差不多的单个模型;在单个模型预测精度相差不大时,单个模型的数量越多,组合模型的精度会越高;单个模型的拟合精度越高,对应的组合模型的拟合精度也越高;组合模型相对误差与单个模型的相对误差的偏离方向和大小有关;在相同预测精度要求下,组合模型需要的观测时间比单个模型要短。

Using the data from observing the subgraae sememem o~ u,~ ~ Railway, the variable-weight combination forecasting for settlement of loess subgrade is analyzed and studied. The result shows that when the single model is selected according to the fitting accuracy, those having reverse deviation direction of relative error and litde difference on the magnitude should be selected by all means; when the forecasting precision with the single model has little discrepancy, the greater the number of the single model is, the higher accuracy the combination model has ; and the higher fitting accuracy the single model has, and so has the corresponding combination model; the relative error of the combination model relates to the deviation direction and magnitude of the relative error of the single model; based on the same forecasting accuracy requirement, the observation time required by the combination model is shorter than that of the single model.