Periodic Table’s Properties Using Unsupervised Chemometric Methods: Undergraduate Analytical Chemistry Laboratory Exercise

An application of unsupervised chemometrics methods using periodic table properties for an analytical chemistry laboratory exercise is presented. Chemometric techniques such as hierarchical clustering analysis (HCA), k-means, and principal component analysis (PCA) were applied to a multivariate data set of chemical properties of the elements (atomic radius, electronegativity, ionization energy, electronic affinity, thermal conductivity, density, entropy, and specific heat). The exercise was carried out by undergraduate students attending a chemometric analysis class during the fifth semester of their third year at our educational institution. The theory of HCA, k-means, and PCA is discussed, and multivariate analysis procedures (data set construction, preprocessing, dendrogram, k-means clustering, scores, and loadings) were carried out using the well-liked programming language R, a widely used programming language designed for data analysis and statistics, within the user-friendly RStudio integrated development environment. The unsupervised algorithms were able to find “natural” clustering from the periodic table using the data structure without any prior knowledge of the class assignment of the samples. This analytical chemistry laboratory exercise with chemometric techniques can also be used in a wide range of laboratory activities such as water analysis, food analysis, drug analysis, and physical chemistry.本文介绍了一种将元素周期表属性应用于分析化学实验课程的无监督化学计量学方法案例。研究将层次聚类分析（hierarchical clustering analysis, HCA）、k均值聚类（k-means）以及主成分分析（principal component analysis, PCA）等化学计量学技术，应用于元素化学属性的多变量数据集，该数据集涵盖原子半径、电负性、电离能、电子亲和能、热导率、密度、熵与比热容等指标。该实验由我校三年级第五学期修读化学计量分析课程的本科生完成。本文阐述了HCA、k均值聚类与PCA的理论基础，并借助广受欢迎的数据分析与统计专用编程语言R，在操作便捷的RStudio集成开发环境中完成了多变量分析全流程，包括数据集构建、数据预处理、聚类树状图绘制、k均值聚类、主成分得分计算与主成分载荷分析等步骤。上述无监督算法无需预先知晓样本的类别分配信息，仅依靠数据结构即可从元素周期表的属性数据中挖掘出"自然"聚类结果。这套结合化学计量学技术的分析化学实验课程，还可推广应用于水质分析、食品分析、药物分析以及物理化学等众多实验活动中。