Study on the Recognition of Liquor Age of Gujing Based on Raman Spectra and Support Vector Regression
WANG Guo-xiang1, WANG Hai-yan2, WANG Hu1, ZHANG Zheng-yong2, LIU Jun1
1. School of Management Science & Engineering, Nanjing University of Finance & Economics，Nanjing 210046, China 2. Jiangsu Province Institute of Quality Safety & Engineering, Nanjing 210000, China
Abstract：It is an important and difficult research point to recognize the age of Chinese liquor rapidly and exactly in the field of liquor analyzing, which is also of great significance to the healthy development of the liquor industry and protection of the legitimate rights and interests of consumers. Spectroscopy together with the pattern recognition technology is a preferred method of achieving rapid identification of wine quality, in which the Raman Spectroscopy is promising because of its little affection of water and little or free of sample pretreatment. So, in this paper, Raman spectra and support vector regression (SVR) are used to recognize different ages and different storing time of the liquor of the same age. The innovation of this paper is mainly reflected in the following three aspects. First, the application of Raman in the area of liquor analysis is rarely reported till now. Second, the concentration of studying the recognition of wine age, while most studies focus on studying specific components of liquor and studies together with the pattern recognition method focus more on the identification of brands or different types of base wine. The third one is the application of regression analysis framework, which cannot be only used to identify different years of liquor, but also can be used to analyze different storing time, which has theoretical and practical significance to the research and quality control of liquor. Three kinds of experiments are conducted in this paper. Firstly, SVR is used to recognize different ages of 5, 8, 16 and 26 years of the Gujing Liquor; secondly, SVR is also used to classify the storing time of the 8-years liquor; thirdly, certain group of train data is deleted form the train set and put into the test set to simulate the actual situation of liquor age recognition. Results show that the SVR model has good train and predict performance in these experiments, and it has better performance than other non-liner regression method such as the Partial Least Squares Regression (PLS) method, and can also be applied in the practice of liquor analysis.
Key words：Raman spectra;Support vector regression;Recognition of wine age;Recognition of wine storing time
王国祥1，王海燕2，王 虎1，张正勇2，刘 军1 . 基于Raman光谱和支持向量机回归的古井贡酒年份鉴别方法 [J]. 光谱学与光谱分析, 2016, 36(03): 729-735.
WANG Guo-xiang1, WANG Hai-yan2, WANG Hu1, ZHANG Zheng-yong2, LIU Jun1 . Study on the Recognition of Liquor Age of Gujing Based on Raman Spectra and Support Vector Regression. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2016, 36(03): 729-735.
[1] Yu H, Lin H, Xu H, et al. Journal of Agricultural and Food Chemistry, 2008, 56(2): 307. [2] Riu-Aumatell M, Bosch-Fusté J, López-Tamames E, et al. Food chemistry, 2006, 95(2): 237. [3] Pinho O, Ferreira I M, Santos L H. Journal of Chromatography A, 2006, 1121(2): 145. [4] Soufleros E H, Bouloumpasi E, Tsarchopoulos C, et al. Food Chemistry, 2003, 80(2): 261. [5] ZHUANG Ming-yang(庄明扬). Sichuan Food and Fermentation(四川食品与发酵)，2008, 44(3)：28. [6] XU Zhan-cheng(徐占成). Sichuan Food and Fermentation(四川食品与发酵)，2008, 44(2)：9. [7] WANG Li, WANG Di-qiang, WANG Hua, et al(王 莉，汪地强，汪 华，等). Liquor-making(酿酒)，2005, 32(4)：18. [8] Vapnik V N. The Nature of Statistical Learning Theory. New York: Springer-Verlag, 1995. [9] Hsu C W, Chang C C, Lin C J. Practical Guide to Support Rector Classification, 2003.
