三维荧光光谱结合PARAFAC和GA对中国白酒品牌的鉴别

Identification of Chinese Liquors by Three-Dimensional Fluorescence Spectra Combined with PARAFAC and Genetic Algorithm

为了实现对中国白酒品牌的有效鉴别,在比较了不同品牌白酒三维荧光光谱特性的基础上,采用平行因子方法(PARAFAC)结合遗传算法(GA)获得训练样本和测试样本的浓度得分。同时利用支持向量机(SVM)方法建立中国白酒的鉴别模型,预测准确率为97.5%。PARAFAC分解得到的三个组分的浓度得分,在一定程度上反映了品牌之间的差异。PARAFAC与GA算法的有效结合为未知样本的鉴别提供了一种快速准确的方法。研究结果表明,PARAFAC-GASVM的鉴别模型具有更强的预测能力,该方法能够有效提取白酒的特征光谱信息,同时又大大降低了SVM输入变量的维数。研究结果为中国白酒的鉴别提供了一种新的思路。

In order to classify the brands of Chinese liquors effectively, the fluorescence characters of Chinese liquors are compared and analyzed. The concentration scores of training samples and testing samples are obtained by using parallel factor method （PARAFAC） combined with genetic algorithm （GA）. Support vector machine （SVM） method is adopted to establish the identification model of Chinese liquors, and the accuracy rate of prediction is 97.5%. The experimental results show that the concentration scores of the three principal components reflect the difference between brands. The combination of PARAFAC and GA provides an accurate method for the rapid identification of unknown samples. The results indicate that PARAFAC-GA-SVM has higher prediction accuracy. The proposed method can effectively extract the spectral characteristics, and also reduce the dimension number of the input variables of SVM. The results can provide a new way for the identification of Chinese liquors.