ARIMA与GM(1,1)模型在某深基坑沉降预测中的对比分析

Comparative Analysis of ARIMA and GM(1,1) Model in Deep Foundation Pit Settlement Prediction

准确预测深基坑支护变形对提高基坑工程安全性具有重要意义。该文以武汉华中科创产业园超大深基坑工程实际沉降数据为依据,采用SPSS与MATLAB软件分别建立ARIMA与GM模型,并对实际沉降数据进行拟合效果分析。结果表明:1)GM(1,1)模型拟合结果线性趋势过大,不利于基坑变形的后期预测;2)ARIMA(0,2,0)模型拟合结果与原始序列趋势基本一致,其MSE值与相对残差均小于GM(1,1)模型,更适用于该工程基坑变形预测研究;3)为避免ARIMA(0,2,0)模型随期数增加而样本量减少,导致后期拟合数据残差增大的缺点,需要及时补充新数据。

Accurately predicting the deformation of deep foundation pit support has great significance to improve the safety of foundation pit engineering.Based on the actual settlement data of the super deep foundation pit project of Wuhan Huazhong Science and Technology Innovation Industrial Park,the ARIMA model and GM model are established by SPSS and MATLAB software respectively,and the fitting effect of the actual settlement data is analyzed.The results show that:1)The linear trend of GM(1,1)model fitting result is too large,which is not conducive to the later prediction of foundation pit deformation;2)The fitting results of ARIMA(0,2,0)model are basically consistent with the original sequence trend,and the MSE value and relative residual error are all smaller than GM(1,1)model,which is more suitable for the prediction of foundation pit deformation in this project;3)In order to avoid the disadvantage that the sample size of the ARIMA(0,2,0)model decreases with the increase of the number of periods,resulting in the increase of residual error of later fitting data,it is necessary to supplement the new data in time.