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

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 350¹800nm. 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.