葡萄酒中元素分布与其原产地关系的分类模型

为了解决进口葡萄酒来源复杂，原产地难鉴别的问题，使用电感耦合等离子体质谱检测葡萄酒中的元素含 量，采用偏最小二乘法建立聚类分类模型用于原产地鉴别。电感耦合等离子体质谱检测了澳大利亚、智利、法国、 意大利和西班牙5 个国家的100 份葡萄酒中的41 种元素，通过变量两两相乘进行扩维，偏最小二乘法变量筛选方法 对扩维后的大量变量进行处理，删除冗余变量和影响不显著的变量，建立了聚类分析模型，模型可以很好地将各国 葡萄酒样品区分，分辨准确率在96%以上。将来自5 个国家的99 份和南非的11 份葡萄酒样品检测数据代入模型，判别结果令人满意。

A Classification Model for Wines Based on Elements Distribution and Geographical Origin

In this study, efforts were made to address the difficulty in identifying the geographic origin of imported wines to
China due to the complex sources. Inductively coupled plasma mass spectrometry (ICP-MS) was used to detect the contents
of 41 elements in 100 samples of imported wines from Australia, Chile, France, Italy and Spain. Furthermore, analysis of
the experimental data by partial least squares (PLS) method and cluster analysis was carried out to establish a classification
model for tracing the geographical origin of the wines. After variable dimension expansion and variable selection, the PLS
model without redundant or non-significant variables showed a good correlation coefficient of cross validation with an
identification accuracy above 96%. For 110 additional new wine samples, including 99 from the five countries and 11 from
South Africa, satisfactory identification results were obtained when applying the analytical data to the model.