Effective multivariate data presentation and modeling in distinction of the tea infusions

Abstract

In this work a review of multivariate data visualization and unsupervised modeling methods was presented, applied to DPV and LAPV signals of black, green, red and white tea infusions of China and Vietnam origin. It proves that graphical presentation by diagrams and charts has a number of advantages in comparison to analysis of the measured signals or numerical datasheets. The same data presented and modeled by different methods may deliver various, complementary conclusions and also the strategy verifies the quality of the experiments. Insight to the experimental data and extraction of the useful information were done applying Chernoff faces, star plots, radar plot and matrix scatterplot. Using SOMs and PCA it was demonstrated that DPV data enabled distinction of the tea kinds with some problems with separation of the green and white ones, while LAPV signals distinguished the origin of the infusions. Successful separation of the 8 different types of objects were done by polar dendrogram, first time applied in voltammetry.