New tools for exploratory analysis fusing information from different sources

Research output: Book/ReportPh.D. thesis

  • Nicola Cavallini
The main focus of this PhD project was on the development and application of new chemometric tools for multivariate exploratory analysis for dealing with data not showing simple groupings or trends, even when projected to spaces of lower dimensionality. Such data may be so complex that common visualizations tools are only shedding limited light on the underlying structures. Starting from these premises, the Fused Adjacency Matrix approach was developed as main outcome of the project. The approach was tested on a benchmark dataset of beer samples acquired using three spectroscopic techniques, namely visible, near-infrared (NIR) and nuclear magnetic resonance (NMR) spectroscopies. Another important part of the PhD project concerned the extension of methods for integration of data sources of very different nature, like numerical and text data, within the food chemistry framework. As a matter of fact, analytical chemistry in synergy with advanced data analysis methods can be profitably used to build new tools to aid consumers to choose and pair foodstuff as well as producers to meet the consumers’ expectations and desires. In this perspective, an investigation of the links between the “objective” world of analytical chemical profiling and the “subjective” world of consumers tasting and describing food was carried out, in the context of beer analysis and consumption. By means of text analysis methods, a set of user-generated reviews were processed and converted into a numeric format suitable for data analysis, and then linked by principal component analysis–generalized canonical analysis (PCA–GCA) to the spectral information provided by Visible, NIR and NMR spectroscopies.
Original languageEnglish
PublisherDepartment of Food Science, Faculty of Science, University of Copenhagen
Number of pages211
Publication statusPublished - 2019

ID: 249063999