最小角回归结合竞争性自适应重加权采样的近红外光谱波长选择

doi:10.3964/j.issn.1000-0593(2021)06-1782-07

Abstract
References
Related Citation (15)

Download: PDF (2102 KB)
Export: BibTeX | EndNote (RIS)

Abstract Near-infrared spectroscopy is widely used in drug detection, petrochemical industry, etc., because it has no damage to the samples, and the detection speed is fast, and the accuracy is high. In particular, it has more accurate detection performance with the in-depth application of machine learning and deep learning modeling methods in recent years makes. However, the NIR spectral data of the sample has relatively high dimensions and has problems such as spectral overlap, collinearity and noise, which will negatively impact the performance of the NIR spectral model. In this case, the selection of effective characteristic wavelength points of the sample is extremely important. In order to improve the accuracy and reliability of the quantitative and qualitative analysis models of NIR spectra, a variable selection method for NIR spectra is proposed, which combines the advantages of Least Angle Regression and Competitive Adaptive Re-weighted Sampling, and has better performance. In this method, LAR was used to preliminarily screen the characteristic wavelengths in the whole spectrum of the sample, and then CARS was used to further select the selected characteristic wavelengths to effectively remove the irrelevant characteristic wavelengths. In order to verify the effectiveness of the method, the method was evaluated from two aspects of quantitative and qualitative analysis. In the quantitative analysis experiment, PLS regression analysis model was established using FULL, LAR, CARS, SPA and UVE as comparison methods and drug sample data set as example. PLS model established by variables screened by LAR-CARS showed higher predictive determination coefficient and lower predictive standard deviation in drug data set. In the qualitative analysis experiment, the classification model was established with SVM, ELM, SWELM and BP as comparison methods and drug data sets with different proportions of training sets. The accuracy of the SVM classification model established by the variables screened by LAR-CARS reached the highest 100%. From the experimental results, it can be seen that LAR-CARS can effectively select the wavelength points that the characteristics of the sample, and the quantitative and qualitative analysis model established by using the selected wavelength points has better robustness and can be used for the characteristic wavelength screening of the sample spectrum.

Key words：Near infrared spectroscopy; Wavelength selection; Least angle regression; Competitive adaptive reweighted sampling

Received: 2020-07-01 Accepted: 2020-10-28

ZTFLH:

O657.3

Corresponding Authors: LI Ling-qiao E-mail: 54pe@163.com

Cite this article:

URL:

https://www.gpxygpfx.com/EN/10.3964/j.issn.1000-0593(2021)06-1782-07 OR https://www.gpxygpfx.com/EN/Y2021/V41/I06/1782

[1] Yazici A, Tiryaki G Y, Ayvaz H, et al. Journal of the Science of Food and Agriculture, 2020, 100(5): 1980.
[2] Greene T P, Gullysantiago M A, Barsony M. The Astrophysical Journal, 2018, 862(1): 85.
[3] Li S, Xing B, Lin D, et al. Industrial Crops and Products, 2020, 152(11): 112539.
[4] ZHANG Feng, TANG Xiao-jun, TONG Ang-xin, et al(张峰，汤晓君，仝昂鑫，等). Chinese Journal of Scientific Instrument(仪器仪表学报), 2020, 41(1): 64.
[5] Felipe C B, Espinosa R M, Hevia G, et al. Journal of Sports Sciences, 2019, 37(23): 1.
[6] Liu J, Zhang Y, Wang H, et al. Spectrochimica Acta Part A: Molecular & Biomolecular Spectroscopy, 2018, 199(21): 43.
[7] Wang Y, Guo W, Zhu X, et al. International Journal of Food Science & Technology, 2019, 54(2): 387.
[8] Tsakiridis N L, Tziolas N V, Theocharis J B, et al. European Journal of Soil Science, 2019, 70(3): 578.
[9] WANG Kun, WU Jing-zhu, WANG Dong, et al(王坤，吴静珠，王冬，等). Journal of Food Safety and Quality Detection Technology(食品安全质量检测学报)，2020，11(16)：5569.
[10] LI Xin-xing, YAO Jiu-bin, CHENG Jian-hong, et al(李鑫星, 姚久彬, 成建红, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2020, 40(1): 189.
[11] ZHAO Huan, HUAN Ke-wei, SHI Xiao-guang, et al(赵环, 宦克为, 石晓光, 等). Chinese Journal of Analytical Chemistry(分析化学), 2018, 65(1): 136.
[12] Chen F C, Jahanshahi M R. IEEE Transactions on Industrial Electronics, 2018, 65(5): 4392.
[13] Zhang J, Lu Z, Li M, et al. IEEE Access, 2019, 7(1): 183118.
[14] YANG Zhen-fa, XIAO Hang, ZHANG Lei, et al(杨振发，肖航，张雷，等). Chinese Journal of Analytical Chemistry(分析化学)，2020，48(2)：275.
[15] Krongchai C, Wongsaipun S, Funsueb S, et al. Chiang Mai Journal of Science, 2020, 41(1): 160.
[16] Lu B, Liu N, Li H, et al. Soil & Tillage Research, 2019, 191(12): 266.
[17] Zhang R, Zhang F, Chen W, et al. Chemometrics & Intelligent Laboratory Systems, 2018, 175(11): 47.