Network analysis on Fourier-transform infrared (FTIR) spectroscopic data sets in an Eigen space layout: Introducing a novel approach for analysing wine samples

Abstract

In the present work, Eigen-directed network analysis for Fourier-transform infrared (FTIR) spectroscopic data sets of wine samples was introduced. A network can generally be viewed as a collection of nodes connected to each other through links, often also called edges. Herein, each node in the network represents a sample and the dissimilarity weight associated with the difference between the two connected nodes is described by the edge. The utility of the approach was tested by analysing a collection of 148 wine samples. The networking on FTIR data sets of these samples in the Eigen space layout was found to impart required aesthetic values as well as the chemical significance to the nodes positioning. The proposed approach successfully captured the compositional differences among the analysed wine samples and classified them in two groups. The Eigen-directed network analysis also allowed a swift assessment regarding inter- and intra-group homogeneity. Homogeneous groups were found to contain nodes with high degree of adjacency and edges with smaller lengths. In comparative study, the proposed approach was found to outperform the network analysis in force-directed layout and principal component analysis. In summary, the proposed Eigen-directed network analysis provided a simplified illustration of highly correlated spectral data sets enabling a swift and intuitive interpretation.

Keywords: Degree of adjacency; Eigen-directed layout; Homogeneity; Network; Wine.