zhpsЖурнал прикладной спектроскопииZhurnal Prikladnoii Spektroskopii0514-7506B. I. Stepanov Institute of Physics of the National Academy of Scienceszhps-1500Research ArticleАННОТАЦИИ АНГЛОЯЗЫЧНЫХ СТАТЕЙABSTRACTS ENGLISH-LANGUAGE ARTICLESСравнительное исследование калибровочных моделей с использованием данных ИК-спектроскопии в ближнем диапазонеComparative Study on Calibration Models Using NIR Spectroscopy DataPanN.PanN.

Школа математики 

School of Mathematics and Big Data

ning.cecil.pan@outlook.comYuZ.YuZ.

Школа математики 

School of Mathematics and Big Data

LingW.LingW.

Школа математики 

School of Mathematics and Big Data

XuJ.XuJ.

Школа математики 

School of Mathematics and Big Data

LiaoY.LiaoY.

Школа математики 

School of Mathematics and Big Data

Университет образования ГуйчжоуGuizhou Education University202414022024911172172Copyright © Pan N., Yu Z., Ling W., Xu J., Liao Y., 20242024Pan N., Yu Z., Ling W., Xu J., Liao Y.Pan N., Yu Z., Ling W., Xu J., Liao Y.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://zhps.ejournal.by/jour/article/view/1500

Для определения влажности и содержания жира и белка в свинине разработана автоматическая процедура, основанная на регрессии опорных векторов (SVR), методе обратного распространения ошибки (BPNN) и объединении анализа главных компонент и BPNN (PCA-BPNN) с использованием 16 комбинаций предварительной обработки (функция-свертка, скользящее среднее, удаление тренда на основе стандартной нормальной переменной и мультипликативной коррекции разброса). Сравнение моделей проведено для оценки влияния предварительной обработки и калибровки на прогнозирующую способность моделей. Методы коррекции и сглаживания позволяют значительно уменьшить ошибку прогнозирования модели. Большинство моделей SVR имеют высокую точность прогнозирования и подходят для прогнозирования влажности и содержания белка. BPNN и PCABPNN больше подходят для устранения нелинейности между содержанием жира и данными ИК-спектроскопии в ближнем диапазоне.

 

The quality of pork is largely influenced by moisture, fat, and protein. In the meat industry, the establishment of a fast and accurate prediction system is always welcomed. Near infrared spectroscopy (NIRS) can satisfy the requirements of the evaluation. An automatic routine based on support vector regression (SVR), a backpropagation neural network (BPNN), and principal component analysis-backpropagation neural network (PCA-BPNN) was developed to predict three components of pork using 16 combinations of pretreatment (convolution function-based moving average, detrending based on the standard normal variate, and multiplicative scatter correction). Model comparisons were implemented to evaluate the influence of pretreatment and calibration models on the prediction ability of models. The correction method and smoothing methods can significantly reduce the model prediction error. Most of the SVR models have high prediction accuracy and are suitable for predicting moisture and protein. The BPNN and PCA-BPNN are more suitable for dealing with nonlinearity between fat and NIR observations.

жирные кислотыинфракрасная спектроскопия в ближнем диапазонерегрессия опорных векторовметод обратного распространения ошибкианализ главных компонентfatty acidsnear-infrared spectroscopysupport vector regressionback-propagation neural networkprincipal component analysisThis work was supported by the Fund of Guizhou Education Department Youth Science and Technology Talent Growth Project (QianJiaoHe-KY-Zi[2022]315, QianJiaoJi-KY-Zi[2022]263), Science Research Foundation of Guizhou Education University (2022YB008), and the National Natural Science Foundation of China (12001131).ReferencesE. Zamora-Rojas, A. Garrido-Varo, E. De Pedro-Sanz, J. E. Guerrero-Ginel, D. Pérez-Marín, Food Chem., 129, 1889–1897 (2011).E. Zamora-Rojas, A. Garrido-Varo, E. De Pedro-Sanz, J. E. Guerrero-Ginel, D. Pérez-Marín, Food Chem., 129, 1889–1897 (2011).M. Bonneau, B. Lebret, Meat Science, 84, 293–300 (2010).M. Bonneau, B. Lebret, Meat Science, 84, 293–300 (2010).I. Ramírez-Morales, D. Rivero, E. Fernández-Blanco, A. Pazos, Chemometr. Intell. Lab. Syst., 159, 45–57 (2016).I. Ramírez-Morales, D. Rivero, E. Fernández-Blanco, A. Pazos, Chemometr. Intell. Lab. Syst., 159, 45–57 (2016).O. Chapelle, P. Haffner, V. N. Vapnik, IEEE Transact. Neural Networks, 10, 1055–1064 (1999).O. Chapelle, P. Haffner, V. N. Vapnik, IEEE Transact. Neural Networks, 10, 1055–1064 (1999).C. De Bleye, P. F. Chavez, J. Mantanus, R. Marini, P. Hubert, E. Rozet, E. Ziemons, J. Pharm. Biomed. Analysis, 69, 125–132 (2012).C. De Bleye, P. F. Chavez, J. Mantanus, R. Marini, P. Hubert, E. Rozet, E. Ziemons, J. Pharm. Biomed. Analysis, 69, 125–132 (2012).Y. Roggo, L. Duponchel, B. Noe, J. P. Huvenne, J. Near Infrared Spectrosc., 10, 137–150 (2002).Y. Roggo, L. Duponchel, B. Noe, J. P. Huvenne, J. Near Infrared Spectrosc., 10, 137–150 (2002).http://lib.stat.cmu.edu/datasets/tecator.http://lib.stat.cmu.edu/datasets/tecator.C. Borggaard, H. H. Thodberg, Anal. Chem., 64, 545–551 (1992).C. Borggaard, H. H. Thodberg, Anal. Chem., 64, 545–551 (1992).M. B. Whitfield, M. S. Chinn, J. Near Infrared Spectrosc., 25, 363–380 (2017).M. B. Whitfield, M. S. Chinn, J. Near Infrared Spectrosc., 25, 363–380 (2017).X. Chu, H. Yuan, W. Lu, Progress Chem., 16, 528–542 (2004).X. Chu, H. Yuan, W. Lu, Progress Chem., 16, 528–542 (2004).T. Naes, T. Isaksson, B. Kowalski, Anal. Chem., 62, 664–673 (1990).T. Naes, T. Isaksson, B. Kowalski, Anal. Chem., 62, 664–673 (1990).A. Savitzky, M. J. Golay, Anal. Chem., 36, 1627–1639 (1964).A. Savitzky, M. J. Golay, Anal. Chem., 36, 1627–1639 (1964).S. R. Delwiche, R. A. Graybosch, Appl. Spectrosc., 57, 1517–1527 (2003).S. R. Delwiche, R. A. Graybosch, Appl. Spectrosc., 57, 1517–1527 (2003).Q. Liu, J. Wang, IEEE Transact. Neural Networks, 19, 558–570 (2008).Q. Liu, J. Wang, IEEE Transact. Neural Networks, 19, 558–570 (2008).S. Heo, J. H. Lee, Comput. Chem. Eng., 127, 1–10 (2019).S. Heo, J. H. Lee, Comput. Chem. Eng., 127, 1–10 (2019).Y. Wang, Y. Li, Y. Song, X. Rong, Appl. Sci., 10, 1897 (2020).Y. Wang, Y. Li, Y. Song, X. Rong, Appl. Sci., 10, 1897 (2020).S. W. Lin, S. C. Chen, W. J. Wu, et al., Knowledge and Inform. Systems, 21, 249–266 (2009).S. W. Lin, S. C. Chen, W. J. Wu, et al., Knowledge and Inform. Systems, 21, 249–266 (2009).C. H. Li, S. C. Park, Inform. Proc. Manag., 45, 329–340 (2009).C. H. Li, S. C. Park, Inform. Proc. Manag., 45, 329–340 (2009).L. Zhang, Y. Li, Y. Gu, et al., China Comm., 14, 141–150 (2017).L. Zhang, Y. Li, Y. Gu, et al., China Comm., 14, 141–150 (2017).C. Alcaraz, A. Vila‐Gispert, E. García‐Berthou, Diversity and Distributions, 11, 289–298 (2005).C. Alcaraz, A. Vila‐Gispert, E. García‐Berthou, Diversity and Distributions, 11, 289–298 (2005).R. M. Balabin, E. I. Lomakina, Analyst, 136, 1703–1712 (2011).R. M. Balabin, E. I. Lomakina, Analyst, 136, 1703–1712 (2011).H. T. Lin, C. J. Lin, Neural Comput., 3, 1–32 (2003).H. T. Lin, C. J. Lin, Neural Comput., 3, 1–32 (2003).S. Arlot, A. Celisse, Statistics Surv., 4, 40–79 (2010).S. Arlot, A. Celisse, Statistics Surv., 4, 40–79 (2010).J. Li, S. Zhu, S. Jiang, J. Wang, LWT – Food Sci. Technol., 82, 369–376 (2017).J. Li, S. Zhu, S. Jiang, J. Wang, LWT – Food Sci. Technol., 82, 369–376 (2017).P. Shan, S. Peng, Y. Bi, L. Tang, C. Yang, Q. Xie, C. Li, Chemometr. Intell. Lab. Syst., 138, 72–83 (2014).P. Shan, S. Peng, Y. Bi, L. Tang, C. Yang, Q. Xie, C. Li, Chemometr. Intell. Lab. Syst., 138, 72–83 (2014).

The authors declare that there are no conflicts of interest present.