摘要
利用衰减全反射傅里叶变换红外光谱(ATR-FTIR)技术对聚谷氨酸(γ-PGA)发酵批次进行快速识别检测.根据5个批次的γ-PGA发酵液光谱建立了5个偏最小二乘判别分析(PLSDA)分类模型来鉴别每一批次与其他批次,在某些局部评价指标(准确率)上取得了不错的结果.为了提高模型在全体指标(如准确率、精度、灵敏度等)上的整体效果和模型的可解释性,采用子窗口置换分析(SPA)、竞争自适应重加权采样(CARS)以及随机青蛙(RF)三种波数选择算法结合PLSDA提取特征波数,建立PLSDA判别模型.实验结果表明,经波数选择(除CARS)后,PLSDA在各个批次上的各项性能指标均得到提升,而且模型复杂度降低,可解释性增强;ATR-FTIR技术结合SPA-PLSDA或RF-PLSDA方法可实现对γ-PGA发酵批次的快速识别.
Attenuated total reflection Fourier transform infrared spectroscopy(ATR-FTIR)technology is utilized to quickly identify and detect different fermentation batches forγ-polyglutamic-acid(γ-PGA).Based on partial least squares discriminant analysis(PLSDA),five classification models were established to distinguish each batch from other batches,which performed well on several individual evaluation indicators(e.g.,accuracy).In order to improve the model performance on all the indicators(e.g.,accuracy,precision,sensitivity,etc.)and the model interpretability,three wavenumber selection methods including subwindow permutation analysis(SPA),competitive adaptive reweighted sampling(CARS)and random frog(RF)were combined with PLSDA to extract key wavenumbers and then established the corresponding classification models.Experimental results show that all the performance indicators of PLSDA combined with wavenumber selection(except CARS-PLSDA)are improved.Additionally,both the model complexity and interpretability are improved.Therefore,ATR-FTIR technology combined with SPA-PLSDA or RF-PLSDA method can realize rapid identification of differentγ-PGA fermentation batches.
作者
单鹏
吴缀
何年
刘隆兴
SHAN Peng;WU Zhui;HE Nian;LIU Long-xing(School of Control Engineering,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China)
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2022年第10期1376-1382,共7页
Journal of Northeastern University(Natural Science)
基金
中央高校基本科研业务费专项资金资助项目(N2023021).
关键词
衰减全反射
聚谷氨酸
偏最小二乘判别分析
竞争自适应
波数选择
attenuated total reflection
γ-polyglutamic-acid(γ-PGA)
partial least squares discriminant analysis(PLSDA)
competitive adaptive
wavenumber selection
