基于数量化理论Ⅲ的滑坡变形组合预测研究

[目的]探讨基于数量化理论Ⅲ的组合预测模型在滑坡变形组合中的适用性,为滑坡变形组合预测提供新的思路。[方法]选取GM(1,1)、支持向量机、BP神经网络及MC预测作为滑坡变形的单一预测模型,并利用数量化理论Ⅲ分析单项预测的相对误差区间的重要性,进而确定区间权重,再结合预测误差确定的组间权重,综合确定组合预测的权值。[结果]组合预测结果相对误差的平均值为1.1%,方差为0.16。[结论]组合预测较单一预测模型的预测精度及稳定性都有较大的提高。

Forecasting Deformation Combination of Landslide Based on Third Theory of Quantification

[Objective] To explore the applicability of the combined forecasting model based on the third theory of quantification in combined landslide deformation and to provide a new idea for forecasting the combined landslide deformation. [Methods] We used GM(1, 1), support vector machine, BP neural network and MC prediction as a single prediction model of landslide deformation, and applied the third theory of quantification to analyze the importance of relative error interval and single prediction, and then determine the interval weights and the prediction error between groups, thus determining the combination weights of comprehensive prediction. [Results] The average value of relative error of combined prediction was 1.1% and the variance was 0.16. [Conclusion] Compare to single forecast model, the forecast accuracy and stability of combined prediction had been greatly improved.