可见/近红外结合MIA变量优选和支持向量机判别山茶油的制取方式

Discrimination of extraction method of camellia oil by Vis/NIR spectra and MIA-SVM

山茶油的主要制取方式有压榨法和浸出法,且压榨山茶油的品质优于浸出法。本研究利用可见/近红外光谱技术结合化学计量学对山茶油的制取方式进行判别研究。采集不同制取方式的山茶油样本在350-1800nm波段范围的可见/近红外光谱,利用边界影响分析（margin influence analysis,MIA）新方法进行波长变量优选,并应用支持向量机（support vector machines,SVM）对优选的波长变量建立山茶油制取方式的判别分类模型。结果表明：可见/近红外光谱联合MIA-SVM方法判别山茶油的制取方式是可行的,其校正集和预测集样本的灵敏度、特异性及正确率分别为100%、87.50%、93.75和100%、87.50%、93.75%。说明MIA是一种有效的波长变量选择方法,能简化分类模型,提高分类模型的稳定性和预测精度。

The main ways of camellia oil production are pressing and extraction ,and the quality of pressed camellia oil is superior to extracted camellia oil. The objective of this research was to discriminate extraction method of camellia oil by visible/near infrared（Vis/NIR） spectroscopy. Spectra of camellia oil samples extracted by different methods were acquired in the wavelength range of 350N1800nm. Margin influence analysis（MIA） was used to select sensitive wavelength variables,then support vector machines （SVM） was used to develop classification model for extraction method of camellia oil using the selected wavelength variables. The results indicate that Vis/NIR spectra combined with MIA-SVM method is feasible for discriminating extraction method of camellia oil,the sensitivity,specificity and accuracy of classification model for samples in calibration and prediction sets are 100%, 87.50%, 93.75% and 100%, 87.50%, 93.75%, respectively. This demonstrates that MIA was an effective variable selection method, it could simplify classification model, improved stability and prediction accuracy of classification model.