ResearchArticle
Comparison of Machine Learning Classification Methods for
Determining the Geographical Origin of Raw Milk Using
Vibrational Spectroscopy
Aimen El Orche ,
1
Amine Mamad,
2
Omar Elhamdaoui,
2
Amine Cheikh,
3
Miloud El Karbane,
2
and Mustapha Bouatia
2
1
Team of Analytical and Computational Chemistry,Nanotechnology and Environment, Faculty of Sciences and Techniques,
University of Sultan Moulay Slimane, Beni Mellal, Morocco
2
Laboratory of Analytical Chemistry, Faculty of Medicine and Pharmacy, Mohammed V University, Rabat, Morocco
3
Faculty of Medicine, Abulcasis University, Rabat, Morocco
Correspondence should be addressed to Aimen El Orche; aimen.elorche@gmail.com
Received 1 October 2021; Revised 22 November 2021; Accepted 26 November 2021; Published 8 December 2021
Academic Editor: Ana Domi nguez Vidal
Copyright © 2021 Aimen El Orche et al. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
One of the significant challenges in the food industry is the determination of the geographical origin, since products from different
regions can lead to great variance in raw milk. erefore, monitoring the origin of raw milk has become very relevant for
producers and consumers worldwide. In this exploratory study, midinfrared spectroscopy combined with machine learning
classification methods was investigated as a rapid and nondestructive method for the classification of milk according to its
geographical origin. e curse of dimensionality makes some classification methods struggle to train efficient models. us,
principal component analysis (PCA) has been applied to create a smaller set of features. e application of machine learning
methods such as PLS-DA, PCA-LDA, SVM, and PCA-SVM demonstrates that the best results are obtained using PLS-DA, PCA-
LDA, and PCA-SVM methods which show a correct classification rate (CCR) of 100% for PLS-DA and PCA-LDA and 94.95% for
PCA-SVM, whereas the application of SVM without feature extraction gives a low CCR of 66.67%. ese findings demonstrate
that FT-MIR spectroscopy, combined with machine learning methods, is an efficient and suitable approach to classify the
geographical origins of raw milk.
1. Introduction
Consumers are increasingly demanding guarantees con-
cerning the quality and safety of food products, especially
when they are of animal origin. Food authentication consists
of checking that the product is consistent with the state-
ments made on the label [1]. Falsification or willful mis-
labeling is usually used to reduce production costs [2]. In the
milk industry especially, mislabeling can be used to confuse
the industry and consumers about the origin of the milk,
since products of different origins can have different qual-
ities [1].
e European Union (EU) promotes dairy product
quality programs designed to support farmers and safeguard
their product names from abuse and imitation [3]. Partic-
ularly, the EU promotes two principal quality regimes that
are based on the valuation of the geographical origins of
foods, called Protected Designation of Origin (PDO) and
Protected Geographical Indication (PGI), which respectively
identify food products that are produced or closely asso-
ciated with a given geographical area [4].
Food scientists support such programs by developing
analytical techniques to enhance the capacity to identify the
geographic origin of food [5]. ese techniques are classified
into two types: those based on targeted approaches and those
based on untargeted approaches [6].
Targeted approaches are usually the most suitable for
regulatory purposes, as they are very specific and sensitive
Hindawi
Journal of Spectroscopy
Volume 2021, Article ID 5845422, 9 pages
https://doi.org/10.1155/2021/5845422