摘要
针对传统回归模型在高层建筑物沉降预测中,出现的由于自变量的选择而造成的多重共线等问题,本文提出了一种以FPE准则定阶的时间序列(AR)模型。以桂林市临桂县某农贸市场周边建筑物的沉降变形为例,在分析利用自相关函数和偏相关函数来对模型进行判定和识别且用FPE定阶准则确定模型的阶数的基础上,对建筑物的沉降观测数据进行建模分析,得到了均方误差为0.055 6的预测精度,很好地克服了传统回归模型中存在的问题,结论具有一定的实用价值。
In order to cope with multiple collinearity problems of traditional regression model,due to the inde-pendent variables selection in the subsidence prediction for tall buildings,a time sequence (AR)model based on FPE to set the order is proposed.The subsidence of a farmers market in Guilin is the case in our research. The autocorrelation function and partial correlation function are used to determine and identify the model,and FPE order criteria is also used to determine the model order.On the basis of modeling analysis and observed buildings data,in a high predicted accuracy the mean square error is 0.055 6,and overcomes the problems in traditional regression model with a practical value.
出处
《桂林理工大学学报》
CAS
北大核心
2014年第1期85-88,共4页
Journal of Guilin University of Technology
基金
国家自然科学基金项目(41071294)
关键词
时间序列
AR模型
建筑物沉降
预测
time series
AR model
subsidence of buildings
prediction