摘要
基于分位点回归模型比最小二乘回归模型具有更强的统计分析能力,将参数估计的MM算法和统计诊断中的影响度量MM距离运用于大坝水平位移的统计模型,并给出了应用实例.结果表明,分位点回归方法受异常点影响较小,训练及预报结果优于最小二乘法,将其应用于大坝水平位移安全监测,可提高预报模型的预测精度,能更好地对大坝运行性态进行分析.
Based on the superiority of the quantile regression model to the least square regression model in statistical analysis,the MM algorithm of parameters and the MM distance for influence measure in the statistical diagnostics were introduced into the statistical model for horizontal displacement of dams.An application example was presented.The results show that the quantile regression method is less influenced by the abnormal data,and its training and predicted results are superior to those of the least square regression method.The proposed method is feasible for the horizontal displacement of dam safety monitoring and may improve the prediction accuracy of relevant models.It can be employed to analyze the operation behaviors of dams.
出处
《河海大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第1期99-103,共5页
Journal of Hohai University(Natural Sciences)
基金
国家自然科学基金(51079046
50909041
50809025
50879024)
"十一五"国家科技支撑计划(2006BAC14B03
2008BAB29B03)
河海大学水文水资源与水利工程科学国家重点实验室专项基金(2009586012
2010585212)
中央高校基本科研业务费项目(2009B08514
2010B20414
2010B14114
2010B01414)
中国水电工程顾问集团公司科技项目(CHC-KJ-2007-02)
江苏省"333高层次人才培养工程"科研项目(2017-B08037)
江苏省普通高校研究生科研创新计划(CX09B_163Z)
高等学校博士学科点专项科研基金(20070294023)
关键词
非线性分位点回归
MM算法
统计诊断
MM距离
大坝安全监测
nonlinear quantile regression
MM algorithm
statistical diagnostics
MM distance
dam safety monitoring
