Evaluation of the Performance of Multi-Block Data Fusion (Colorimetry, Spectrophotometric Color Indices, Antioxidant Activity) Combined with Machine-Learning for the Characterization of the Geographical Origin of Greek Wines

Read full text Summarise with VIRAT
Authors: Paris Christodoulou, Dimitra Z. Lantzouraki, Panagiotis G. Zoumpoulakis, Dionisis Cavouras, Vassilia J. Sinanoglou
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Assessment of the geographical origin of wine is important in checking the quality of the products and protecting the regulatory provisions. Therefore, the current study suggests a low-cost, multi-block, data-fusion approach that combines the Commission Internationale de l’Éclairage (CIE) Lab* (CIELAB) color parameters, spectrophotometric color indices and antioxidant/antiradical activities to categorize wines produced in four Greek regions, including Aegean Islands, Crete, Macedonia and Peloponnese. Patterns that were found to be region-specific were revealed through exploratory principal component analysis and were mainly influenced by chromatic and pigment-related variables. A full modeling strategy utilizing linear classifiers (Linear Discriminant Analysis (LDA) and Partial Least Squares Discriminant Analysis (PLS-DA)) and non-linear ensemble approaches (Random Forest, Support Vector Machine with radial basis function kernel (SVM-RBF) and Gradient Boosting) was employed and assessed using out-of-bag bootstrap estimation and fivefold cross-validation. Random Forest consistently achieved the best predictive performance, with 0.900 out-of-bag and 0.865 cross-validated accuracy using the integrated dataset, thereby outperforming any single-block model. Visualization of proximity-maps defined chemically coherent clusters such as a tightly clustered region of the Aegean Islands, and a partial adjacency between continental regions. Analysis of feature importance showed that the most significant classifiers were absorbance at 520 and 420 nm, color density, hue, and other CIELAB parameters. Together, the results demonstrate that multi-block fusion datasets, offers a strong and cost-effective platform on which to perform routine wine-origin characterization.

Fields of Study & Subjects

Spectroscopy and Chemometric Analyses Horticultural and Viticultural Research Fermentation and Sensory Analysis

Publication details

Journal / source
Figshare
Publication date
28 March 2026
Language
EN