基于提升建模的锌离子与钴离子浓度紫外可见吸收光谱检测方法
|
摘要
紫外可见分光光度法检测高浓度Zn离子和痕量Co(Ⅱ)离子混合溶液时,由于Zn(Ⅱ)离子对痕量Co(Ⅱ)离子吸收光谱的掩蔽,以及两种离子之间化学性质相近,经常导致光谱重叠、相互干扰。针对这一问题,本研究提出一种基于提升建模的Zn(Ⅱ)离子和Co(Ⅱ)离子浓度紫外可见吸收光谱检测方法。本方法通过对校正集加权采样获得子数据集;然后使用子数据集建立不同压缩比的LASSO回归子模型集,使用赤池信息量准则(AIC)选择最优子模型;根据子模型对建模样本的误差大小,更新样本权重,重复迭代建模至子模型收敛;最后根据子模型的预测性能给予子模型不同的权重,加权融合子模型得到最终的总模型。共获得80组Zn(Ⅱ)离子和Co(Ⅱ)离子混合溶液的紫外可见光谱数据集,将本方法与全波段的偏最小二乘(PLS)、蒙特卡洛无信息变量消除(MCUVE)-PLS及竞争自适应重加权采样(CARS)-PLS进行了比较分析,对于Zn(Ⅱ)离子,本方法保留的有效波长点个数相比PLS、MCUVE-PLS和CARS-PLS都大幅减少,预测均方根误差相对于PLS、MCUVEPLS和CARS-PLS分别减少55. 3%、21. 3%和1. 64%。对于Co(Ⅱ)离子,本方法保留的有效波长点个数相比MCUVE-PLS和CARS-PLS大量减少,降低了模型的复杂度,预测均方根误差相对于PLS、MCUVE-PLS和CARS-PLS分别减少71.4%、46.2%和54.8%。
Ultraviolet-visible( UV-vis) absorption spectrophotometry is used to detect the concentration of mixed solution of Zn(Ⅱ) and Co(Ⅱ) ions,where the problems of spectrum overlaps and interferes with each other due to their similar chemical properties are well solved. A boosting modeling method based on least absolute shrinkage and selection operator( LASSO) regression is proposed. The method first obtains sub-data sets by weighting the calibration samples,and then uses the sub-data sets to establish LASSO sub-model sets with different penalty factors,and uses Akaike Information Criterion( AIC) to judge the sub-model sets. According to the error of sub-model to modeling sample,the sample weight is updated. The iterations are repeated until sub-model convergence. Finally,the sub-models are given different weights according to the prediction performance of the sub-model,and the final prediction model is obtained by weighting fusion of the sub-model.A total of 80 groups of UV-Vis spectral data sets of mixed solution of Zn (Ⅱ) and Co (Ⅱ) are obtained by experiments. The method is compared with PLS of the whole spectrum,Monte Carlo uniformative variable elimination-partial least square( MCUVE-PLS) and competitive adaptive reweighted sampling( CARS)-PLS,and it is found that that the proposed method can greatly retain the wavelength variables with high contribution to the model and improve both the explanatory and predictive performance of the model. The method can be used to establish a stable and efficient prediction model for spectral data analysis.
Ultraviolet-visible( UV-vis) absorption spectrophotometry is used to detect the concentration of mixed solution of Zn(Ⅱ) and Co(Ⅱ) ions,where the problems of spectrum overlaps and interferes with each other due to their similar chemical properties are well solved. A boosting modeling method based on least absolute shrinkage and selection operator( LASSO) regression is proposed. The method first obtains sub-data sets by weighting the calibration samples,and then uses the sub-data sets to establish LASSO sub-model sets with different penalty factors,and uses Akaike Information Criterion( AIC) to judge the sub-model sets. According to the error of sub-model to modeling sample,the sample weight is updated. The iterations are repeated until sub-model convergence. Finally,the sub-models are given different weights according to the prediction performance of the sub-model,and the final prediction model is obtained by weighting fusion of the sub-model.A total of 80 groups of UV-Vis spectral data sets of mixed solution of Zn (Ⅱ) and Co (Ⅱ) are obtained by experiments. The method is compared with PLS of the whole spectrum,Monte Carlo uniformative variable elimination-partial least square( MCUVE-PLS) and competitive adaptive reweighted sampling( CARS)-PLS,and it is found that that the proposed method can greatly retain the wavelength variables with high contribution to the model and improve both the explanatory and predictive performance of the model. The method can be used to establish a stable and efficient prediction model for spectral data analysis.
引文
1 ZHU Hong-Qiu, GONG Juan, LI Yong-Gang, CHEN Jun-Ming. Spectroscopy and Spectral Analysis, 2017, 37(12): 3882-3888朱红求, 龚娟, 李勇刚, 陈俊名. 光谱学与光谱分析, 2017, 37(12): 3882-3888
2 ZHU Hong-Qiu, CHEN Jun-Ming, YIN Dong-Hang, LI Yong-Gang, YANG Chun-Hua. Journal of Chemical Industry, 2017, 68(03): 998-1004朱红求, 陈俊名, 尹冬航, 李勇刚,阳春华. 化工学报, 2017, 68(03): 998-1004
3 LIANG Yi-Zeng, WU Hai-Long, YU Ru-Qing. Handbook of Analytical Chemistry·Chemometrics. Beijing: Chemical Industry Press, 2016: 211-212梁逸曾, 吴海龙, 俞汝勤. 分析化学手册·化学计量学. 北京: 化学工业出版社, 2016: 211-212
4 Poerio D V, Brown S D. Chemometr. Intell. Lab. Sys., 2017, 166(2017): 49-60
5 Blanchet F G, Legendre P, Borcard D. Ecology, 2008, 89(9): 2623-2632
6 Sutter J M, Kalivas J H. Microchem. J., 1993, 47(1): 60-66
7 Centner V, Massart D, deNoord O E, deJong S, Vandeginste B M, Sterna C. Anal. Chem., 1996, 68(21): 3851-3858
8 Yun Y H, Wang W T, Tan M L, Liang Y Z, Li H D. Anal. Chim. Acta, 2014, 807(1): 36-43
9 Sun X D, Zhou M X, Sun Y Z. Infrared Phys. Tech., 2016, 77(1): 65-72
10 Li H D, Liang Y Z, Xu Q S. Anal. Chim. Acta, 2009, 648(1): 77-84
11 Colombani C, Legarra A, Fritz S, Guillaume F, Croiseau P, Ducrocq V. J. Dairy Sci., 2013, 96(1): 575-591
12 Lê C K A, Rossouw D, Robertgranié C, Besse P. Stat. Appl. Genet. Mol., 2008, 7(1): 11-35
13 Li H D, Liang Y Z, Xu Q S. Chemometr. Intell. Lab. Sys., 2010, 104(2): 341-346
14 Bian X, Li S, Shao X G, Liu P. Chemometr. Intell. Lab. Sys., 2016, 158(1): 174-179
15 Li Z G, Lv J T, Si G Y. Chemometr. Intell. Lab. Sys., 2015, 146(1): 211-220
16 Breiman L. Mach. Learn., 1996, 24(2): 123-140
17 Zou P C, Wang J, Chen S, Chen H. Knowl. Based Sys., 2014, 65(1): 21-30
18 Cao D S, Xu Q S, Liang Y Z, Zhang L X, Li H D. Chemometr. Intell. Lab. Sys., 2010, 99(1): 1-11
19 Gao F, Kou P, Gao L, Guan X. Neurocomputing, 2013, 113(7): 67-87
20 Shao X, Bian X, Cai W. Anal. Chim. Acta, 2010, 666(1): 32-37
21 Inouye T, Toi S, Matsumoto Y. Cognitive Brain Res., 1995, 3(1): 33-40
22 Macho S, Rius A, Callao M P, Larrechi M S. Anal. Chim. Acta, 2001, 445(2): 213-220
23 Martens H, Naes T. J. Chromatogr. A, 2007, 33(3): 366-367
2 ZHU Hong-Qiu, CHEN Jun-Ming, YIN Dong-Hang, LI Yong-Gang, YANG Chun-Hua. Journal of Chemical Industry, 2017, 68(03): 998-1004朱红求, 陈俊名, 尹冬航, 李勇刚,阳春华. 化工学报, 2017, 68(03): 998-1004
3 LIANG Yi-Zeng, WU Hai-Long, YU Ru-Qing. Handbook of Analytical Chemistry·Chemometrics. Beijing: Chemical Industry Press, 2016: 211-212梁逸曾, 吴海龙, 俞汝勤. 分析化学手册·化学计量学. 北京: 化学工业出版社, 2016: 211-212
4 Poerio D V, Brown S D. Chemometr. Intell. Lab. Sys., 2017, 166(2017): 49-60
5 Blanchet F G, Legendre P, Borcard D. Ecology, 2008, 89(9): 2623-2632
6 Sutter J M, Kalivas J H. Microchem. J., 1993, 47(1): 60-66
7 Centner V, Massart D, deNoord O E, deJong S, Vandeginste B M, Sterna C. Anal. Chem., 1996, 68(21): 3851-3858
8 Yun Y H, Wang W T, Tan M L, Liang Y Z, Li H D. Anal. Chim. Acta, 2014, 807(1): 36-43
9 Sun X D, Zhou M X, Sun Y Z. Infrared Phys. Tech., 2016, 77(1): 65-72
10 Li H D, Liang Y Z, Xu Q S. Anal. Chim. Acta, 2009, 648(1): 77-84
11 Colombani C, Legarra A, Fritz S, Guillaume F, Croiseau P, Ducrocq V. J. Dairy Sci., 2013, 96(1): 575-591
12 Lê C K A, Rossouw D, Robertgranié C, Besse P. Stat. Appl. Genet. Mol., 2008, 7(1): 11-35
13 Li H D, Liang Y Z, Xu Q S. Chemometr. Intell. Lab. Sys., 2010, 104(2): 341-346
14 Bian X, Li S, Shao X G, Liu P. Chemometr. Intell. Lab. Sys., 2016, 158(1): 174-179
15 Li Z G, Lv J T, Si G Y. Chemometr. Intell. Lab. Sys., 2015, 146(1): 211-220
16 Breiman L. Mach. Learn., 1996, 24(2): 123-140
17 Zou P C, Wang J, Chen S, Chen H. Knowl. Based Sys., 2014, 65(1): 21-30
18 Cao D S, Xu Q S, Liang Y Z, Zhang L X, Li H D. Chemometr. Intell. Lab. Sys., 2010, 99(1): 1-11
19 Gao F, Kou P, Gao L, Guan X. Neurocomputing, 2013, 113(7): 67-87
20 Shao X, Bian X, Cai W. Anal. Chim. Acta, 2010, 666(1): 32-37
21 Inouye T, Toi S, Matsumoto Y. Cognitive Brain Res., 1995, 3(1): 33-40
22 Macho S, Rius A, Callao M P, Larrechi M S. Anal. Chim. Acta, 2001, 445(2): 213-220
23 Martens H, Naes T. J. Chromatogr. A, 2007, 33(3): 366-367