[1]聂 斌,杜玉文,杜建强*,等.融入距离方差和距离相关系数的偏最小二乘回归方法[J].江西师范大学学报(自然科学版),2023,(01):61-68.
　NIE Bin,DU Yuwen,DU Jianqiang*,et al.The Regression Method of PLS Fusing Distance Variance and Distance Correlation Coefficient[J].Journal of Jiangxi Normal University:Natural Science Edition,2023,(01):61-68.
点击复制

融入距离方差和距离相关系数的偏最小二乘回归方法()
分享到：

《江西师范大学学报》（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2023年01期
页码:
61-68
栏目:
信息科学与技术
出版日期:
2023-01-25

文章信息/Info

Title:
The Regression Method of PLS Fusing Distance Variance and Distance Correlation Coefficient
作者:
聂 斌; 杜玉文; 杜建强*; 张玉超; 郑学鹏; 靳海科
(江西中医药大学计算机学院,江西 南昌 330004)
Author(s):
NIE Bin; DU Yuwen; DU Jianqiang*; ZHANG Yuchao; ZHENG Xuepeng; JIN Haike
(School of Computer,Jiangxi University of Chinese Medicine,Nanchang Jiangxi 330004,China)
关键词:
偏最小二乘;  距离方差;  距离相关系数;  回归方程;  拟线性
Keywords:
partial least square;  distance variance;  distance correlation coefficient;  regression equations;  quasilinear
分类号:
TP 311
文献标志码:
A
摘要:
偏最小二乘法(partial least square,PLS)在内部采用Pearson系数度量自变量和因变量之间的相关性时提取出的成分不能确保解释性最强,并且PLS在将提取的成分进行线性回归时也无法真实反映变量间的函数关系.针对这些问题,该文提出了融入距离方差和距离相关系数的偏最小二乘回归方法(DVDCCPLS).DVDCCPLS基于距离方差和距离相关系数提取距离成分,再将距离成分进行拟线性回归得到距离回归方程,通过模型求解方法将距离回归方程转换为原始数据的表达,最终得到结构简洁、精度较高的回归模型.该文分别采用麻杏石甘汤数据和UCI数据集测试DVDCCPLS的性能,并与其他5种经典的回归算法对比,结果表明:DVDCCPLS具有较好的回归效果和回归性能.
Abstract:
Partial least square(PLS)internally adopts Pearson coefficient to measure the correlation between independent and dependent variables,but the extracted components cannot ensure the strongest interpretation.In addition,PLS applies linear regression to the extracted components,which is unable to reflect the functional relationship between variables truly.Therefore,the regression method of partial least square fusing distance variance and distance correlation coefficient(DVDCCPLS)is proposed to solve the above problems.DVDCCPLS extracts the distance component based on the distance variance and distance correlation coefficient,and then performs quasilinear regression to obtain the distance regression equation.Finally,the distance regression equation is converted into the expression of the original data by the model solution method,and the regression model with simple structure and high precision is obtained in the end.The Maxingshigan decoction datasets and UCI datasets are respectively used to test the performance of DVDCCPLS,and the other five classical regression algorithms are compared with DVDCCPLS.The results show that DVDCCPLS has better regression effects and performances.

参考文献/References:

[1] GELADI P,KOWALSKI B R.Partial least squares regression:a tutorial [J].Analytica Chimica Acta,1986,185:1-17.
[2] YOU Xinge,MOU Yi,YU Shujian,et al.Mixed-norm partial least squares [J].Chemometrics and Intelligent Laboratory Systems,2016,152:42-53.
[3] MALTHOUSE E C,TAMHANE A C,MAH R S H.Nonlinear partial least squares [J].Computers & Chemical Engineering,1997,21(8):875-890.
[4] LIU Hongbin,YANG Chong,BENGT C,et al.Dynamic nonlinear partial least squares modeling using Gaussian process regression [J].Industrial & Engineering Chemistry Research,2019,58(36):16676-16686.
[5] 尚栋,孙兰香,齐立峰,等.基于循环变量筛选非线性偏最小二乘的LIBS铁矿浆定量分析 [J].中国激光,2021,48(21):171-179.
[6] MA Hao,WANG Yan,JI Zhicheng.A novel dynamic nonlinear partial least squares based on the cascade structure [J].International Journal of Robust and Nonlinear Control,2022,32(6):3584-3605.
[7] 贾润达,毛志忠,王福利.基于KPLS模型的间歇过程产品质量控制 [J].化工学报,2013,64(4):1332-1339.
[8] JIAO Jianfang,ZHAO Ning,WANG Guang,et al.A nonlinear quality-related fault detection approach based on modified kernel partial least squares [J].ISA Transactions,2016,66:275-283.
[9] ZHU Bao,CHEN Zhongsheng,HE Yanlin,et al.A novel nonlinear functional expansion based PLS(FEPLS)and its soft sensor application [J].Chemometrics and Intelligent Laboratory Systems,2017,161:108-117.
[10] WANG Yanxia,CAO Hui,ZHOU Yan,et al.Nonlinear partial least squares regressions for spectral quantitative analysis [J].Chemometrics and Intelligent Laboratory Systems,2015,148:32-50.
[11] 鲁庆华,任康乐,周凤玺.基于偏最小二乘法实现非线性回归分析 [J].甘肃科技,2005,21(11):146-148.
[12] MERINO A,GARCIA-ALVAREZ D,SAINZ-PALMERO G,et al.Knowledge based recursive non-linear partial least squares(RNPLS)[J].ISA Transactions,2020,100:481-494.
[13] 李雄威,郭晓雅,李庚达,等.一种基于非线性偏最小二乘的风电机组齿轮箱状态监测方法 [J].可再生能源,2022,40(10):1346-1351.
[14] LAVOIE F B,MUTEKI K,GOSSELIN R.A novel robust NL-PLS regression methodology [J].Chemometrics and Intelligent Laboratory Systems,2018,184:71-81.
[15] 谢文龙.三次样条函数的构造方法 [J].江南学院学报,2000,15(2):90-93.
[16] INDAHL U.A twist to partial least squares regression [J].Journal of Chemometrics,2005,19(1):32-44.
[17] LAVOIE F B,MUTEKI K,GOSSELIN R.Generalization of powered-partial-least-squares [J].Chemometrics and Intelligent Laboratory Systems,2018,179:1-11.
[18] PENG Shan,PENG Silong,TANG Liang,et al.A nonlinear partial least squares with slice transform based piecewise linear inner relation [J].Chemometrics and Intelligent Laboratory Systems,2015,143:97-110.
[19] LI Fachao,YANG Kuo.Research of the regression method based on quasi-linear function [EB/OL].[2022-08-13].https://ieeexplore.ieee.org/document/5662989/.
[20] WOLD S,RUHE A,WOLD H,et al.The collinearity problem in linear regression:the partial least squares(PLS)approach to generalized inverses [J].SIAM Journal on Scientific and Statistical Computing,1984,5(3):735-743.
[21] SZÉKELY G J,RIZZO M L,BAKIROV N K.Measuring and testing dependence by correlation of distances [J].Annals of Statistics,2007,35(6):2769-2794.

备注/Memo

备注/Memo:
收稿日期:2022-11-02
基金项目:国家自然科学基金(82260849,62141202),民族药资源数据库与信息网络化共享平台构建(2019YFC1712301)和江西中医药大学校级科技创新团队发展计划(CXTD22015)资助项目.
作者简介:聂 斌(1972—), 男,江西峡江人,教授,博士研究生,主要从事数据挖掘、中医药信息学和中药学的研究.E-mail:ncunb@163.com
通信作者:杜建强(1968—),男,江西南昌人,教授,博士,博士导师,主要从事中医药信息与数据挖掘的研究.E-mail:jianqiang_du@163.com
更新日期/Last Update: 2023-01-25