基于变量频次加权自助采样法的近红外光谱变量选择方法研究

引用本文: 陈笑, 宦克为, 赵环, 范恒晔, 韩雪艳. 基于变量频次加权自助采样法的近红外光谱变量选择方法研究[J]. 分析化学, 2021, 49(10): 1743-1749. doi: 10.19756/j.issn.0253-3820.210456 shu

Citation: CHEN Xiao, HUAN Ke-Wei, ZHAO Huan, FAN Heng-Ye, HAN Xue-Yan. Variable Selection of Near Infrared Spectroscopy Based on Variable Frequency Weighted Bootstrap Sampling[J]. Chinese Journal of Analytical Chemistry, 2021, 49(10): 1743-1749. doi: 10.19756/j.issn.0253-3820.210456 shu

基于变量频次加权自助采样法的近红外光谱变量选择方法研究

长春理工大学理学院, 长春 130022

通讯作者: 宦克为,E-mail:huankewei@126.com

基金项目:
吉林省科技发展计划项目（No.20190701024GH）资助。

摘要: 近红外光谱分析技术已被广泛应用于食品检测及定量分析等领域。变量选择作为近红外光谱建模分析中的关键步骤，对于提高模型的稳定性和预测性能具有重要作用。本研究提出了一种近红外光谱变量选择方法，即变量频次加权自助采样法（Variable frequency weighted bootstrap sampling，FWBS），通过二进制矩阵采样（BMS）法随机组合生成不同变量子集，利用偏最小二乘法（PLS）建立不同变量子集的子模型，统计每个变量在子模型中出现的频率，并进行归一化处理，得到变量的初始权重，最后结合加权自助采样法（WBS）选取最佳特征变量。以玉米和牛奶近红外光谱公共数据为例，构建了玉米中淀粉含量预测模型和牛奶中蛋白质含量预测模型。结果表明，FWBS-PLS预测模型与无信息变量消除结合PLS （UVE-PLS）、竞争性自适应重加权采样结合PLS （CARS-PLS）和自助软收缩结合PLS （BOSS-PLS）预测模型相比，在玉米数据集上的预测均方根误差（RMSEP）分别由0.2523、0.1162和0.0831下降到0.0740，预测精度分别提升了70.7%、36.3%和11.0%，FWBS-PLS预测模型的相对分析误差（RPD）为11.1328；在牛奶数据集上的RMSEP分别由0.1743、0.1437和0.1432下降到0.0887，预测精度分别提升了49.1%、38.3%和38.1%，FWBS-PLS预测模型的RPD为12.2701。基于变量频次加权自助采样法的近红外光谱变量选择方法可以极大程度地简化模型，提高模型的预测精度。

关键词:

English

Variable Selection of Near Infrared Spectroscopy Based on Variable Frequency Weighted Bootstrap Sampling

School of Science, Changchun University of Science and Technology, Changchun 130022, China

Corresponding author: HUAN Ke-Wei, huankewei@126.com

Received Date: 22 April 2021
Revised Date: 28 July 2021
Fund Project:
Supported by the Jilin Province Science and Technology Development Plan Project (No.20190701024GH).

Abstract: Near-infrared spectroscopy (NIRS) has been widely used in various fields such as food detection and quantitative analysis. As a key step in NIRS modeling and analysis, variable selection plays an important role in improving the stability and predictive performance of model. A new variable selection method for NIRS, variable frequency weighted bootstrap sampling (FWBS), was proposed in this work. Different variable subsets were generated by random combination of binary matrix sampling (BMS), and sub-models of different variable subsets were established by partial least squares (PLS). The frequency of each variable in the sub-model was counted and normalized to get the initial weight of the variable. Finally, the best characteristic variables were selected by weighted bootstrap sampling (WBS). Taking the public data base of NIRS spectroscopy of corn and milk as examples, the prediction models of starch content in corn and protein content in milk were established. The results showed that, comparing with uninformative variable elimination by PLS (UVE-PLS), competitive adapative reweighted sampling by PLS (CARS-PLS) and bootstrapping soft shrinkage by PLS (BOSS-PLS), the root mean square error of prediction (RMSEP) of the FWBS-PLS on the corn dataset decreased from 0.2523, 0.1162 and 0.0831 to 0.0740, respectively. The prediction accuracy increased by 70.7%, 36.3% and 11.0%, respectively. The relative percent deviation (RPD) of FWBS-PLS was 11.1328. The RMSEP on the milk dataset decreased from 0.1743, 0.1437 and 0.1432 to 0.0887, respectively. The prediction accuracy increased by 49.1%, 38.3% and 38.1%, respectively. The RPD of FWBS-PLS was 12.2701. The variable selection of NIRS based on FWBS could simplify model and improve the prediction accuracy of model greatly.

Key words:

1. [1]
  YUN Yong-Huan, DENG Bai-Chuan, LIANG Yi-Zeng. Chin. J. Anal. Chem., 2015, 43(11):1638-1647. 云永欢, 邓百川, 梁逸曾. 分析化学, 2015, 43(11):1638-1647.
2. [2]
  CHU Gang-Hui, WANG Kun, YIN Xue-Bo. Chin. J. Anal. Chem., 2020, 48(4):536-542. 楚刚辉, 王坤, 尹学博. 分析化学, 2020, 48(4):536-542.
3. [3]
  BIAN X H, LI S J, SHAO X G, LIU P. Chemom. Intell. Lab. Syst., 2016, 158(10):174-179.BIAN X H, LI S J, SHAO X G, LIU P. Chemom. Intell. Lab. Syst., 2016, 158(10):174-179.
4. [4]
  YUN Y H, LI H D, DENG B C, CAO D S. TrAC-Trends Anal. Chem., 2019, 113:102-115.YUN Y H, LI H D, DENG B C, CAO D S. TrAC-Trends Anal. Chem., 2019, 113:102-115.
5. [5]
  ZHAO Huan, HUAN Ke-Wei, SHI Xiao-Guang, ZHENG Feng, LIU Li-Ying, LIU Wei, ZHAO Chun-Ying. Chin. J. Anal. Chem., 2018, 46(1):136-142. 赵环, 宦克为, 石晓光, 郑峰, 刘丽莹, 刘微, 赵春英. 分析化学, 2018, 46(1):136-142.
6. [6]
  LI Si-Hai, ZHAO Lei. Spectrosc. Spectral Anal., 2019, 39(4):1047-1052. 李四海, 赵磊. 光谱学与光谱分析, 2019, 39(4):1047-1052.
7. [7]
  LI H H, ZHU J J, JIAO T H, WANG B, WEI W Y, ALI S, OUYANG Q, ZUO M, CHEN Q S. Spectrochim. Acta, Part A, 2020(243):118765.LI H H, ZHU J J, JIAO T H, WANG B, WEI W Y, ALI S, OUYANG Q, ZUO M, CHEN Q S. Spectrochim. Acta, Part A, 2020(243):118765.
8. [8]
  LIANG Yi-Zeng, XU Qing-Song. Instrumental Analysis of Complex Systems-White, Gray and Black Analytical Systems and Their Multivariate Methods. Beijing:Chemical Industry Press, 2012:241-265. 梁逸曾, 许青松. 复杂体系仪器分析-白、灰、黑分析体系及其多变量解析方法. 北京:化学工业出版社, 2012:241-265.
9. [9]
  ZHANG J, CUI X Y, CAI W S, SHAO X G. Sci. China:Chem., 2019, 62(2):271-279.ZHANG J, CUI X Y, CAI W S, SHAO X G. Sci. China:Chem., 2019, 62(2):271-279.
10. [10]
  AL-KAF H A G, ALDUAIS N A M, SAAD A H Y, CHIA K S, MOHSEN A M, ALHUSSIAN H, MAHDI A A M H, SALAM W S W. IEEE Access, 2020:3023681.AL-KAF H A G, ALDUAIS N A M, SAAD A H Y, CHIA K S, MOHSEN A M, ALHUSSIAN H, MAHDI A A M H, SALAM W S W. IEEE Access, 2020:3023681.
11. [11]
  FU Dan-Dan, WANG Qiao-Hua, GAO Sheng, MA Mei-Hu. Chin. J. Anal. Chem., 2020, 48(2):289-297. 付丹丹, 王巧华, 高升, 马美湖. 分析化学, 2020, 48(2):289-297.
12. [12]
  YUN Y H, CAO D S, TAN M L, YAN J, REN D B, XU Q S, YU L, LIANG Y Z. Chemom. Intell. Lab. Syst., 2014, 130(2):76-83.YUN Y H, CAO D S, TAN M L, YAN J, REN D B, XU Q S, YU L, LIANG Y Z. Chemom. Intell. Lab. Syst., 2014, 130(2):76-83.
13. [13]
  ZHANG Xiao-Yu, LI Qing-Bo, ZHANG Guang-Jun. Spectrosc. Spectral Anal., 2014, 34(5):1429-1433. 张晓羽, 李庆波, 张广军. 光谱学与光谱分析, 2014, 34(5):1429-1433.
14. [14]
  YUN Y H, WANG W T, TAN M L, LIANG Y Z, LI H D, CAO D S, LU H M, XU Q S. Anal.Chim. Acta, 2014, 807:36-43.YUN Y H, WANG W T, TAN M L, LIANG Y Z, LI H D, CAO D S, LU H M, XU Q S. Anal.Chim. Acta, 2014, 807:36-43.
15. [15]
  DENG B C, YUN Y H, CAO D S, YIN Y L, WANG W T, LU H M, LUO Q Y, LIANG Y Z. Anal. Chim. Acta, 2016, 908:63-74.DENG B C, YUN Y H, CAO D S, YIN Y L, WANG W T, LU H M, LUO Q Y, LIANG Y Z. Anal. Chim. Acta, 2016, 908:63-74.
16. [16]
  YAN H, SONG X Z, TIAN K D, GAO J X, LI QQ, XIONG Y M, MIN S G. Spectrochim. Acta, Part A, 2019, 210:362-371.YAN H, SONG X Z, TIAN K D, GAO J X, LI QQ, XIONG Y M, MIN S G. Spectrochim. Acta, Part A, 2019, 210:362-371.
17. [17]
  YUN Y H, WANG W T, DENG B C, LAI G B, LIU X B, REN D B, LIANG Y Z, FAN W, XU Q S. Anal. Chim. Acta, 2015, 862:14-23.YUN Y H, WANG W T, DENG B C, LAI G B, LIU X B, REN D B, LIANG Y Z, FAN W, XU Q S. Anal. Chim. Acta, 2015, 862:14-23.
18. [18]
  YUN Y H, LI H D, WOOD L R E, FAN W, WANG JJ, CAO D S, XU Q S, LIANG Y Z. Spectrochim. Acta, Part A, 2013, 111:31-36.YUN Y H, LI H D, WOOD L R E, FAN W, WANG JJ, CAO D S, XU Q S, LIANG Y Z. Spectrochim. Acta, Part A, 2013, 111:31-36.
19. [19]
  YAN H, ZHANG J X, GAO J X, HUANG Y M, XIONG Y M, MIN S G. Sci. Rep., 2018, 8:14729.YAN H, ZHANG J X, GAO J X, HUANG Y M, XIONG Y M, MIN S G. Sci. Rep., 2018, 8:14729.
20. [20]
  WANG Chao, WANG Jian-Ming, FENG Mei-Chen, XIAO Lu-Jie, SUN Hui, XIE Yong-Kai, YANG Wu-De. Spectrosc. Spectral Anal., 2018, 38(5):1520-1525. 王超, 王建明, 冯美臣, 肖璐洁, 孙慧, 谢永凯, 杨武德. 光谱学与光谱分析, 2018, 38(5):1520-1525.

扫一扫看文章

计量

PDF下载量: 7
文章访问数: 815
HTML全文浏览量: 73

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

基于变量频次加权自助采样法的近红外光谱变量选择方法研究

通讯作者: 宦克为,E-mail:huankewei@126.com

English

Variable Selection of Near Infrared Spectroscopy Based on Variable Frequency Weighted Bootstrap Sampling

Corresponding author: HUAN Ke-Wei, huankewei@126.com

CCS Chemistry：嵌段共聚物自组装成 “三明治”二维有机材料，实现纳米级压力传感功能

CCS Chemistry：颜徐州、鲍哲南：你的皮肤可能是一片SPMs

CCS Chemistry：工作高效不疲劳，只须让催化剂动起来

CCS Chemistry：静电组装导向聚合- 聚电解质纳米凝胶量化制备新策略

CCS Chemistry：金黄色葡萄球菌醛基脱氢酶是如何识别特异性底物的？

计量

通讯作者: 陈斌, bchen63@163.com

中国化学会秘书处

基于变量频次加权自助采样法的近红外光谱变量选择方法研究

通讯作者: 宦克为,E-mail:huankewei@126.com

English

Variable Selection of Near Infrared Spectroscopy Based on Variable Frequency Weighted Bootstrap Sampling

Corresponding author: HUAN Ke-Wei, huankewei@126.com

CCS Chemistry：嵌段共聚物自组装成 “三明治”二维有机材料，实现纳米级压力传感功能

CCS Chemistry：颜徐州、鲍哲南： 你的皮肤可能是一片SPMs

CCS Chemistry：工作高效不疲劳，只须让催化剂动起来

CCS Chemistry：静电组装导向聚合- 聚电解质纳米凝胶量化制备新策略

CCS Chemistry：金黄色葡萄球菌醛基脱氢酶是如何识别特异性底物的？

计量

文章相关

通讯作者: 陈斌, bchen63@163.com

中国化学会秘书处

CCS Chemistry：颜徐州、鲍哲南：你的皮肤可能是一片SPMs