GM-BP 组合预测模型在基坑沉降分析中的应用

摘要

In order to improve the precision of deformation forecast ,this paper uses the combined model of GM(1 ,1) and BP neural network .The grey GM(1 ,1) model is easy to use ,which can achieve good results in the case of limited sample data ,but the fluctuation regularity in forecasting sequence and mutation prediction ability is not strong .While the neural network modeling process is relatively complex ,which needs more training samples ,but for the data of fluctuation and mutation it has good prediction ability .The combined model contains the advantages of two single models ,using the model to forecast foundation pit deformation .The result indicates this model has the higher precision than traditional single forecast m o del .%为提高变形预测的精度，采用GM （1，1）与BP神经网络组合模型进行预测。灰色GM （1，1）模型使用方便，在样本数据较少的情况下能够取得不错的预测效果，但对预测序列存在规律性波动或突变时的预测能力不强；而神经网络模型建模过程相对复杂，需要较多的训练样本，但对于数据存在规律性波动和突变时有很好的预测能力。组合模型融合两者优点，将其应用于基坑沉降数据预测，结果表明，该模型预测精度优于传统的单一预测模型。