Abstract

We consider the ranking of dry white wines and their quality using sensory evaluation from the standpoint of multidimensional analysis when the test samples are presented as points of a multidimensional space with coordinates corresponding to expert scores. The degree of preference between the wine samples is expressed in terms of the similarity of sensory estimates, evaluated by the geometric distance between them. In addition to ranking wines, this approach allowed us to study their cluster structure and reveal three groups of homogeneous wines based on the results of a set of expert scores given to them. A correlation the cluster structure of wines with their places in the ranked list is found. The consistency of wine ratings based on the proposed chemometric approach, arithmetic mean, and median scores is demonstrated. From the point of view of multivariate analysis, rating wines by the geometric distance is mathematically more reasonable, because it is built in accordance with the Euclidean metric, and the sum of coordinates of points, the calculation of the arithmetic mean is based on, does not correspond to any of the known metrics. Another advantage of the geometric approach is that the use of the arithmetic mean for the final assessment of the tasting results is justified at a great number of experts.

中文翻译：

---

通过对干白葡萄酒质量的感官评价结果对干白葡萄酒进行排名的化学计量（几何）方法

摘要

我们从多维分析的角度考虑使用感官评价对干白葡萄酒的排名及其质量，当测试样本被呈现为多维空间的点时，坐标对应于专家分数。葡萄酒样品之间的偏好程度用感官估计的相似性表示，通过它们之间的几何距离进行评估。除了对葡萄酒进行排名之外，这种方法还使我们能够研究它们的聚类结构，并根据一组专家评分的结果揭示三组同质葡萄酒。发现了葡萄酒的集群结构与其在排名列表中的位置的相关性。证明了基于所提出的化学计量学方法、算术平均值和中位数分数的葡萄酒评级的一致性。从多元分析的角度来看，用几何距离对葡萄酒进行评级在数学上更合理，因为它是按照欧几里德度量建立的，并且点的坐标之和，算术平均值的计算是基于，确实不对应任何已知指标。几何方法的另一个优点是，使用算术平均值对品尝结果进行最终评估得到了大量专家的认可。