基于多化学组分的模式识别法在地理标志食醋真伪鉴别中的应用

Application of multicomposition analysis and pattern recognition in identification of geographical indication vinegar

测定了地理标志食醋(山西老陈醋)中的氨基酸、金属元素、多元醇、有机酸和食醋特征成分四甲基吡嗪等5类成分的含量,并且基于这些化学成分的含量结合经典的化学计量学方法——主成分分析(principal component analysis,PCA)和偏最小二乘分析(partial least squares analysis,PLS),建立了适用于地理标志食醋(山西老陈醋)的鉴别方法。基于多化学组分组合建立模型的方式相比于单类成分构建模型的方式更充分利用了样品的综合特征信息,使得分类效果更明显,并且基于组合模型的PLS模型的预测准确率较高。在变量筛选过程中,使用了Fisher权函数变量选择法、类别权函数变量选择法、PCA距离变量选择法对模型变量进行了优化,基于优化后的变量所建立的模型其性能被证实是令人满意的。研究发现,基于氨基酸、无机元素、多元醇、有机酸、食醋特征成分四甲基吡嗪的组合数据建立的PLS模型,可以作为山西老陈醋真伪鉴别的手段。

An identification model of geographical indication vinegar（ Shanxi extra vinegar） was established using the statistical analysis method including principal component analysis（ PCA） and partial least squares method（ PLS）,and based on contents of multicomposition including amino acids,metallic elements,polyalcohols,organic acids and vinegar characteristic component（ tetramethylpyrazine）. The combined identification model based on multicompositions could much better demonstrate the comprehensive information of vinegar sample compared with the model built with one kind of vinegar component. It was found that the forecast accuracy of the identification model built with PLS method was much better. During the variable selection process,fisher weighting function,category weighting function and PCA distance-based method were used. And the identification model built with the selected variables was satisfactory. It was demonstrated that this identification PLS model established with amino acids,metallic elements,polyalcohols,organic acids and vinegar characteristic component（ tetramethylpyrazine）,could be applied to authenticity identification of Shanxi extra vinegars.