近红外光谱结合偏最小二乘判别分析区分造纸法再造烟叶产品

Classification of Paper-process Reconstituted Tobacco by Near-infrared Spectroscopy and Partical Least Squares Discrimination Analysis

本文采集162个造纸法再造烟叶产品的近红外光谱,结合偏最小二乘判别分析(PLS-DA)建立了再造烟叶产品的分类模型,实现了不同牌号再造烟叶产品的快速分类,并对45个预测集样品的牌号进行了分类预测。所建模型对校正集和预测集的预测正确率分别为100.0%和95.5%,与主成分分析(PCA)相比,PLS-DA对不同牌号再造烟叶产品的分类具有更好的效果。该模型为不同牌号再造烟叶产品分类提供了一种新的快速鉴别分析的方法,同时可初步监测再造烟叶产品的质量稳定性。

In order to rapidly classifying different kinds of paper-process reconstituted tobacco,the classification model of reconstituted tobacco was established by partial least squares discrimination analysis（PLS-DA）with the near-infrared spectra of 162 reconstituted tobacco samples.The prediction was carried out on the 45 samples in the prediction set with the established model.The prediction accuracies of training and prediction sets were 100.0% and 95.5%,respectively.Compared with PCA,PLS-DA was more efficient in paper-process reconstituted tobacco classifying.This model provides a new and rapid discrimination analysis method for the classification of reconstituted tobacco.