基于化学计量学集成架构的橡胶手套的差分拉曼光谱分析
Differential Raman Spectroscopy Analysis of Rubber Gloves Based on a Chemometrics Ensemble Architecture
摘要
建立一种基于差分拉曼光谱(DRS)与化学计量学集成的无损、快速检验橡胶手套的方法.采集51份不同品牌、不同用途的橡胶手套样品的DRS;利用主成分分析-线性判别分析(PCA-LDA)对DRS数据进行特征降维与提取;构建集成分类模型,融合支持向量机(SVM)、随机森林(RF)与极端梯度提升(XGBoost)三种算法,并引入理论可分上限评估数据的分类潜力.结果表明,依据DRS特征峰将样品分为四大类.经PCA-LDA降维后,集成模型在测试集上的分类准确率达92.86%,接近理论可分上限(96.43%).所构建的"差分拉曼光谱-化学计量学集成分类架构"有效融合了光谱特征提取、理论性能评估与多模型集成策略,实现了对橡胶手套种类的高精度、无损鉴别.该方法为法庭科学中的橡胶手套物证溯源提供了系统化、可靠的解决方案.
To establish a non-destructive and rapid method for the identification of rubber gloves using differential Raman spectroscopy(DRS)combined with a chemometrics ensemble approach.DRS data were collected from 51 rubber glove samples of different brands and applications.Principal component analysis-linear discriminant analysis(PCA-LDA)was employed for feature dimension reduction and extraction.An ensemble classification model was constructed by integrating three algorithms,namely support vector machine(SVM),random forest(RF),and extreme gradient boosting(XGBoost),and the theoretical upper bound of classifiability was introduced to evaluate the classification potential of the data.The samples were categorized into four major classes based on their DRS characteristic peaks.After PCA-LDA dimension reduction,the ensemble model achieved a classification accuracy of 92.86%on the test set,approaching the theoretical upper bound of classifiability(96.43%).The proposed"DRS-chemometrics ensemble classification architecture"effectively integrates spectral feature extraction,theoretical performance evaluation,and multi-model ensemble strategies,achieving high-precision and non-destructive identification of rubber glove types.This method provides a systematic and reliable solution for the forensic traceability of rubber glove evidence.
年,卷(期):
2026, 89(9)
机标分类号:
TP391
O657.37
TP181
在线出版日期:
2026年10月19日
基金项目:
202204010931006:食品药品安全防控山西省重点实验室开放课题,