光谱学与光谱分析, 2020, 40 (9): 2809, 网络出版: 2020-11-26  

基于氨基酸的仿刺参产地信息认证方法研究

Study on the Origin Information Authentication Method of Apostichopus Japonicus Based on Amino Acids
吴鹏 1,2李颖 1,2,*刘瑀 2,3陈晨 1,2冉明衢 1,2李亚芳 1,2赵新达 3
作者单位
1 大连海事大学航海学院, 辽宁 大连 116026
2 大连海事大学环境信息研究所, 辽宁 大连 116026
3 大连海事大学环境科学与工程学院, 辽宁 大连 116026
摘要
仿刺参富含多种活性物质, 具有极高的药用价值和经济价值, 是水产行业不可或缺的养殖资源。 不同产地的地理环境与营养结构存在显著差异, 所以仿刺参的生长周期与养殖成本相差巨大。 消费者在购买仿刺参时, 会将产地信息作为选择的首要因素, 因为仿刺参的产地直接反映了食品具有的营养价值。 不同产地仿刺参的价格差距悬殊, 面对利益的诱惑, 产地欺诈事件屡禁不止。 因此, 研究一种准确率高、 稳定性好具有优秀泛化能力的仿刺参产地信息认证方法, 能够有效维护品牌产地从业者与消费者的切身利益。 氨基酸是仿刺参营养富集的主要物质, 通过氨基酸特征能够分析出摄食初级生产者的组成, 可以作为产地信息认证的有效工具。 气相色谱-质谱分析(GC-MS)技术能够产生独特的化学指纹图谱用于产地信息鉴别。 对9个产地的156个仿刺参样品, 进行酸水解、 衍生化和酯化等操作, 通过GC-MS测定出氨基酸含量与氨基酸碳稳定同位素数据。 进行置信水平为95%的图基检验, 并利用箱型图检查数据分布, 筛选出13种氨基酸含量和10种氨基酸碳稳定同位素数据。 主成分分析能够在降低数据维度的同时, 挖掘出有价值的信息, 聚集产地识别特性, 提高运算速度与认证精度。 通过交叉验证, 选取前5个主成分作为氨基酸含量和氨基酸碳稳定同位素模型的输入, 累计贡献率分别为98.727%与95.982%。 为了充分挖掘出隐藏在氨基酸数据背后的价值, 选取了8个家族的12个机器学习方法, 共构建出24个单体分类器, 根据数据自身特征找到最优的认证方法。 应用基于遗传交叉因子改进的粒子群优化算法进行模型参数的优化, 得到性能最佳的单体分类器。 结果表明, 氨基酸碳稳定同位素数据具有更优的产地认证特性, 高斯径向基为核的支持向量机与K邻近算法为最佳的两个分类方法。 最后利用集成学习汇集单体分类器的优势, 构建了一种融合多源数据处理方法的仿刺参产地信息认证方法, 模型的平均准确率为99.67%。 建立了仿刺参产地信息认证系统, 为主管机关监管与消费者维权提供了简单可行的手段, 能够有效监管与防止仿刺参产地欺诈事件的发生, 保障了水产养殖行业的稳健发展。
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(9): 2809. 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[J]. Spectroscopy and Spectral Analysis, 2020, 40(9): 2809.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!