Discrimination of geographical origin of extra-virgin olive oils (EVOOs) is of great importance for legislation and consumers worldwide. The feasibility of a rapid discrimination of four different geographical origins of EVOOs with terahertz spectroscopy system was examined. Different chemometrics including least squares-support vector machines (LS-SVM), back propagation neural network (BPNN) and random forest (RF) combined with principal component analysis (PCA), genetic algorithm (GA) were compared to obtain the best discrimination model. The results demonstrated that there were apparent differences among the four different geographical origins of EVOOs in fatty acid compositions and the absorbance spectra, and an excellent classification (accuracy was 96.25% in prediction set) could be achieved using the LS-SVM method combine with GA. It can be concluded that THz spectroscopy together with chemometrics would be a promising technique to rapid discriminate the geographical origin of EVOOs with high efficiency.

对于世界各地的立法和消费者来说，对特级初榨橄榄油（EVOOs）的地理来源的歧视非常重要。研究了使用太赫兹光谱系统快速区分EVOO的四个不同地理来源的可行性。比较了包括最小二乘支持向量机（LS-SVM），反向传播神经网络（BPNN）和随机森林（RF）以及主成分分析（PCA），遗传算法（GA）在内的不同化学计量学，以获得最佳判别模型。结果表明，EVOOs的四个不同地理来源的脂肪酸组成和吸收光谱之间存在明显差异，并且将LS-SVM方法与GA结合使用可以实现极好的分类（预测集中的准确度为96.25％） 。