Abstract：The apostichopus japonicus is rich in a variety of active substances, has extremely high medicinal value and economicvalue, and it is an indispensable aquaculture resource for the fishery Industry. There are significant differences in the geographical environment and trophic structure of different producing areas, consequently, the growth cycle and culturing cost of the apostichopus japonicus vary greatly. When consumers buy apostichopus japonicus, they will use the origin information as the primary factor of choice, because the origin of the apostichopus japonicus directly reflects the nutritional value of the food. The price gap between apostichopus japonicus from different producing areas is wide. In the face of the temptation of interest, it is difficult to prevent the occurrence of origin fraud incidents completely. Therefore, a method of apostichopus japonicus origin information authentication with high accuracy, good stability and excellent generalization ability is studied, which effectively protects the vital interests of brand origin practitioners and consumers. Amino acids are the main substances in the nutrient enrichment of apostichopus japonicus. The amino acid characteristics can be used to analyze the composition of primary producers, and as an effective tool for origin information authentication of apostichopus japonicus. Gas Chromatography-Mass Spectrometry (GC-MS) technology produces unique chemical fingerprints for identification of origin information. The 156 samples of the apostichopus japonicus from 9 producing areas were subjected to acid hydrolysis, derivatization and esterification, and finally, the amino acids content and amino acids carbon stable isotope data were determined by GC-MS. Perform a Tukey’s test with a 95% confidence level, and the box-plot were used to check the data distribution, and screen 13 amino acids content and 10 amino acids carbon stable isotope data. Principal component analysis can reduce the data dimension, valuable mine information, aggregate the origin information identification characteristics, and improve the calculation speed and authentication accuracy of the model at the same time. Through cross-validation, the first five principal components were selected as the input of amino acids content and amino acids carbon stable isotope model, and the accumulative contribution rates were 98.727% and 95.982%, respectively. In order to fully exploit the value hidden behind the amino acids data, this paper selected 12 machine learning methods from 8 families, built a total of 24 monomer classifiers, and found the optimal authentication method according to the characteristics of the data itself. The particle swarm optimization algorithm based on genetic crossover factor improvement was used to optimize the model parameters, and the best performance monomer classifier was obtained. The results show that the carbon of the amino acid stable isotope data has better origin authentication characteristics. The support vector machine (Gaussianradial basis as the kernel function) and the k-nearest neighbor algorithms are the best two classification methods. Finally, leverage ensemble learning to bring together the advantages of monomer classifiers, a method for origin information authentication of apostichopus japonicus with fusioning multi-source data processing methods is constructed. The average accuracy of the model is 99.67%. An origin information authentication system for the apostichopus japonicus is established, which provides a simple and feasible mean for the supervision of the competent authorities and consumer rights protection. The occurrence of the apostichopus japonicus origin fraud incidents is effectively prevented and controlled, and the stable and healthy development of the aquaculture industry is ensured.
吴 鹏，李 颖，刘 瑀，陈 晨，冉明衢，李亚芳，赵新达. 基于氨基酸的仿刺参产地信息认证方法研究[J]. 光谱学与光谱分析, 2020, 40(09): 2809-2814.
WU Peng, LI Ying, LIU Yu, CHEN Chen, RAN Ming-qu, LI Ya-fang, ZHAO Xin-da. Study on the Origin Information Authentication Method of Apostichopus Japonicus Based on Amino Acids. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2020, 40(09): 2809-2814.
[1] Ma D Y，Yang H S，Sun L N，et al. Acta Oceanologica Sinica，2014，33(8)：55.
[2] Zhao Y C，Xue C H，Zhang T T，et al. Journal of Agricultural and Food Chemistry，2018，66(28)：7222.
[3] Moyer D C，DeVries J W，Spink J. Food Control，2017，71：358.
[4] WU Peng，LI Ying，LIU Yu，et al(吴 鹏，李 颖，刘 瑀，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析)，2019，39(5)：1604.
[5] Van Ruth S M，Huisman W，Luning P A. Trendsin Food Science & Technology，2017，67(6)：70.
[6] Cantalapiedra-Hijar G，Ortigues-Marty I，Schiphorst A M，et al. Journal of Agricultural and Food Chemistry，2016，64(20)：4058.
[7] Zhao X D，Liu Y，Li Y，et al. Food Control，2018，91：128.
[8] Boschker H T S，Kromkamp J C，Middelburg J J. Limnologyand Oceanography，2005，50(1)：70.
[9] Paolini M，Ziller L，Laursen K H，et al. Journal of Agricultural and Food Chemistry，2015，63(25)：5841.
[10] Persson J，Nasholm T. Physiologia Plantarum，2001，113(3)：352.
[11] Metges C C，Petzke K J，Hennig U. Journal of Mass Spectrometry，1996，31(4)：367.
[12] Fernandez-Delgado M，Cernadas E，Barro S，et al. Journal of Machine Learning Research，2014，15：3133.
[13] Wolpert D H. Neural Computation，1996，8(7)：1341.
[14] Ghamisi P，Benediktsson J A. IEEE Geoscience and Remote Sensing Letters，2015，12(2)：309.
[15] Sun Z B，Song Q B，Zhu X Y，et al. Pattern Recognition，2015，48(5)：1623.
