分析测试学报2025，Vol.44Issue(6)：1096-1106,11.DOI:10.12452/j.fxcsxb.24122328
紫外光谱结合机器学习算法的祛痘类化妆品中4种禁用抗感染类药物快速筛查
Rapid Screening of 4 Banned Substances in Acne-clearing Cosmetics by UV Spectroscopy Combined with Machine Learning Algorithm
摘要
Abstract
A qualitative model for rapid screening of metronidazole,ketoconazole,chloramphenicol and norfloxacin in acne-clearing cosmetics was developed based on ultraviolet spectrum of cosmetics combined with machine learning algorithms.In this study,ultraviolet spectra of 167 batches of acne-clearing cosmetics were collected for model building.The two-dimensional correlation spectroscopy(2D-COS)technique was used for ultraviolet spectra feature band selection,and the effect of each model was compared under 22 spectral preprocessing methods,three machine learning algorithms,and three dataset division ratios.Five-classification qualitative models were established for positive and negative samples containing metronidazole,ketoconazole,chloramphenicol and norfloxacin,re-spectively.The results showed that the ultraviolet spectra of 190-360 nm were selected to be pro-cessed jointly by standard normal variables(SNV)and Savitzky-Golay convolutional smoothing(SG),and the ratio of training set to prediction set division of 7∶3 was chosen to build a qualitative classifi-cation model using the error back propagation(BP)neural network algorithm.The accuracy of the model training set and prediction set can reach 96.58%and 98.00%,respectively,with good predic-tion and generalisation ability.This method can effectively screen and identify the four banned anti-infective drugs in cosmetics quickly and accurately,which not only saves the detection cost and time and improves the detection efficiency,but also helps the on-site rapid inspection and provides a rap-id and intelligent solution for the detection of illegal addition of banned substances in cosmetics.关键词
紫外光谱/化妆品/误差逆传播神经网络/随机森林/支持向量机/二维相关光谱Key words
ultraviolet spectroscopy/cosmetics/error back propagation neural network/random forest/support vector machines/two-dimensional correlation spectroscopy分类
化学化工引用本文复制引用
向健华,芦丽,方方,石心红..紫外光谱结合机器学习算法的祛痘类化妆品中4种禁用抗感染类药物快速筛查[J].分析测试学报,2025,44(6):1096-1106,11.基金项目
江苏省食品药品监督管理局药品监管科研计划项目(202314) （202314）
