摘要
提出了一种基于核密度估计的核偏鲁棒M-回归(kernel partial robust M-regression based on kernel density estimation,KDE-KPRM)方法。以核密度估计加权策略代替原来的M估计加权策略,利用主成分分析技术和核密度函数识别高杠杆点(输入变量空间异常点),利用残差和核密度函数识别高残差点(输出变量空间异常点),无需反复迭代便可以为样本赋予合适权重,有效地提高了建模速率。通过函数仿真和实际工业仿真,证明了所提出的方法比标准的核偏鲁棒M-回归算法有更好的鲁棒性和更高的建模效率。
This paper proposes a kernel partial robust M-regression based on kernel density estimation(KDE-KPRM)method,which adopts the kernel density estimation for weighting strategy instead of the original M-estimation.The principal component analysis and kernel density function(KDF)are used to identify high leverage points(outlier in the space of input variables).And the residuals and KDF are used to identify high residual points(outlier in the space of output variables).Therefore,samples can be given appropriate weights without iteration,and the modelling efficiency can be effectively improved.Finally,function simulation and practical industrial applications illustrate that the KDE-KPRM method has better robustness and higher modeling efficiency than the original KPRM method.
作者
褚菲
王洁
梁涛
代伟
贾润达
CHU Fei;WANG Jie;LIANG Tao;DAI Wei;JIA Runda(School of Information and Control Engineering, China University of Mining and Technology, Xuzhou, Jiangsu 221000, China;College of Information Science and Engineering, Northeastern University, Shenyang 110819, China;State Key Laboratory of Integrated Automation of Process Industry, Northeastern University, Shenyang 110819, China)
出处
《中国科技论文》
CAS
北大核心
2019年第3期273-277,共5页
China Sciencepaper
基金
国家自然科学基金资助项目(61503384,61873049,61603393)
国家煤加工与洁净化工程技术研究中心开放基金课题资助项目(2018NERCCPP-B03)
江苏省研究生科研与实践创新计划项目(SJCX18_0662)
关键词
核密度估计
主成分分析
核偏鲁棒M-回归
离群点
鲁棒估计
kernel density estimation
principal component analysis(PCA)
kernel partial robust M-regression
outlier
robust estimation
