Machine learning chemometric model for Raman spectroscopy based honey quality assessment
Конференцијски прилог (Објављена верзија)
Метаподаци
Приказ свих података о документуАпстракт
INTRODUCTION:
According to Codex Alimentarius (2001), “Honey is the natural sweet substance, produced by honeybees from the nectar of
plants or from secretions of living parts of plants, or excretions of plant-sucking insects on the living parts of plants, which
the bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store and leave in honeycombs to ripen and mature”. Honey is mostly made up of sugars, as well as enzymes, amino acids, organic acids, vitamins,
aromatic compounds, minerals and carotenoids. It contains a lot of flavonoids and phenolic acids, which have a lot of biological effects and functions such as natural antioxidants, anti-inflammatory and antimicrobial properties. Its composition
is particularly variable, depending on its botanical and geographical origins Because of its exclusive flavor and high dietary
value, natural honey is more expensive than other sweeteners. This is the reason why honey is a target of adultera...tion. The
problem is that counterfeiting honey is relatively easy, but detection is difficult. Further, the authenticity of honey is a global
important problem for commercial producers and consumers. Accordingly, a fast and non-destructive method of detecting
counterfeits is needed.
OBJECTIVES:
The aim of this paper is to verify the possibility of Raman spectroscopy and Support Vector Machine (SVM) for classification
of two different honeys and their fake duplicates. For this purpose, meadow and acacia honeys were selected.
METHOD / DESIGN:
Spectra of homemade and counterfeits honey were recorded using XploRA Raman spectrometer (Horiba Jobin Yvon). Raman scattering was excited by laser at a wavelength of 785 nm equipped with a 600 lines/mm grating; spectra were recorded
by applying exposure time 10 s and accumulated from 10 times scans, using 100% filter. Spectral resolution was 3 cm−1 and
autocalibration was done each time before recording of spectra by 520.47 cm−1 line of silicon. In order to assess a possible
sample inhomogeneity, thirty Raman spectra in the region from 200-3400 cm-1 were recorded for each sample. All spectra
were baseline-corrected, normalized and smoothed. After that PCA (Principal component analysis) was conducted and obtained PCs (first two PCs) served as a features for support vector machine (SVM) classification method. Data were divided into
training model (70 %) and training data (30 %). Pre-processing was done by Unscrambler X 10.4 software (CAMO software,
Norway). In order to determine the best shape of the hyperplane and decision boundary, several kernel function were used:
linear, radial basis and polynomial function. The SVM was conducted by Python and Scikit-learn package.
RESULTS:
Support vector machine showed high accuracy in classification of different honey samples. Accordingly, the best discrimination power showed SVM with polynomial function (100%), followed by radial basis (96.67%) and linear (81.82%).
Кључне речи:
honey / adulteration / raman spectroscopy / support vector machineИзвор:
2021, 5-5Финансирање / пројекти:
- Министарство науке, технолошког развоја и иновација Републике Србије, институционално финансирање - 200116 (Универзитет у Београду, Пољопривредни факултет) (RS-MESTD-inst-2020-200116)
- Conservation of European Biodiversity through Exploitation of Traditional Herbal Knowledge for the Development of Innovative Products (EU-H2020-823973)
URI
https://ibsc2021.pmf.uns.ac.rs/ebook-of-abstracts/http://aspace.agrif.bg.ac.rs/handle/123456789/6549
Институција/група
Poljoprivredni fakultetTY - CONF AU - Kolašinac, Stefan M. AU - Pećinar, Ilinka AU - Dajić Stevanović, Zora P. PY - 2021 UR - https://ibsc2021.pmf.uns.ac.rs/ebook-of-abstracts/ UR - http://aspace.agrif.bg.ac.rs/handle/123456789/6549 AB - INTRODUCTION: According to Codex Alimentarius (2001), “Honey is the natural sweet substance, produced by honeybees from the nectar of plants or from secretions of living parts of plants, or excretions of plant-sucking insects on the living parts of plants, which the bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store and leave in honeycombs to ripen and mature”. Honey is mostly made up of sugars, as well as enzymes, amino acids, organic acids, vitamins, aromatic compounds, minerals and carotenoids. It contains a lot of flavonoids and phenolic acids, which have a lot of biological effects and functions such as natural antioxidants, anti-inflammatory and antimicrobial properties. Its composition is particularly variable, depending on its botanical and geographical origins Because of its exclusive flavor and high dietary value, natural honey is more expensive than other sweeteners. This is the reason why honey is a target of adulteration. The problem is that counterfeiting honey is relatively easy, but detection is difficult. Further, the authenticity of honey is a global important problem for commercial producers and consumers. Accordingly, a fast and non-destructive method of detecting counterfeits is needed. OBJECTIVES: The aim of this paper is to verify the possibility of Raman spectroscopy and Support Vector Machine (SVM) for classification of two different honeys and their fake duplicates. For this purpose, meadow and acacia honeys were selected. METHOD / DESIGN: Spectra of homemade and counterfeits honey were recorded using XploRA Raman spectrometer (Horiba Jobin Yvon). Raman scattering was excited by laser at a wavelength of 785 nm equipped with a 600 lines/mm grating; spectra were recorded by applying exposure time 10 s and accumulated from 10 times scans, using 100% filter. Spectral resolution was 3 cm−1 and autocalibration was done each time before recording of spectra by 520.47 cm−1 line of silicon. In order to assess a possible sample inhomogeneity, thirty Raman spectra in the region from 200-3400 cm-1 were recorded for each sample. All spectra were baseline-corrected, normalized and smoothed. After that PCA (Principal component analysis) was conducted and obtained PCs (first two PCs) served as a features for support vector machine (SVM) classification method. Data were divided into training model (70 %) and training data (30 %). Pre-processing was done by Unscrambler X 10.4 software (CAMO software, Norway). In order to determine the best shape of the hyperplane and decision boundary, several kernel function were used: linear, radial basis and polynomial function. The SVM was conducted by Python and Scikit-learn package. RESULTS: Support vector machine showed high accuracy in classification of different honey samples. Accordingly, the best discrimination power showed SVM with polynomial function (100%), followed by radial basis (96.67%) and linear (81.82%). T1 - Machine learning chemometric model for Raman spectroscopy based honey quality assessment EP - 5 SP - 5 UR - https://hdl.handle.net/21.15107/rcub_agrospace_6549 ER -
@conference{ author = "Kolašinac, Stefan M. and Pećinar, Ilinka and Dajić Stevanović, Zora P.", year = "2021", abstract = "INTRODUCTION: According to Codex Alimentarius (2001), “Honey is the natural sweet substance, produced by honeybees from the nectar of plants or from secretions of living parts of plants, or excretions of plant-sucking insects on the living parts of plants, which the bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store and leave in honeycombs to ripen and mature”. Honey is mostly made up of sugars, as well as enzymes, amino acids, organic acids, vitamins, aromatic compounds, minerals and carotenoids. It contains a lot of flavonoids and phenolic acids, which have a lot of biological effects and functions such as natural antioxidants, anti-inflammatory and antimicrobial properties. Its composition is particularly variable, depending on its botanical and geographical origins Because of its exclusive flavor and high dietary value, natural honey is more expensive than other sweeteners. This is the reason why honey is a target of adulteration. The problem is that counterfeiting honey is relatively easy, but detection is difficult. Further, the authenticity of honey is a global important problem for commercial producers and consumers. Accordingly, a fast and non-destructive method of detecting counterfeits is needed. OBJECTIVES: The aim of this paper is to verify the possibility of Raman spectroscopy and Support Vector Machine (SVM) for classification of two different honeys and their fake duplicates. For this purpose, meadow and acacia honeys were selected. METHOD / DESIGN: Spectra of homemade and counterfeits honey were recorded using XploRA Raman spectrometer (Horiba Jobin Yvon). Raman scattering was excited by laser at a wavelength of 785 nm equipped with a 600 lines/mm grating; spectra were recorded by applying exposure time 10 s and accumulated from 10 times scans, using 100% filter. Spectral resolution was 3 cm−1 and autocalibration was done each time before recording of spectra by 520.47 cm−1 line of silicon. In order to assess a possible sample inhomogeneity, thirty Raman spectra in the region from 200-3400 cm-1 were recorded for each sample. All spectra were baseline-corrected, normalized and smoothed. After that PCA (Principal component analysis) was conducted and obtained PCs (first two PCs) served as a features for support vector machine (SVM) classification method. Data were divided into training model (70 %) and training data (30 %). Pre-processing was done by Unscrambler X 10.4 software (CAMO software, Norway). In order to determine the best shape of the hyperplane and decision boundary, several kernel function were used: linear, radial basis and polynomial function. The SVM was conducted by Python and Scikit-learn package. RESULTS: Support vector machine showed high accuracy in classification of different honey samples. Accordingly, the best discrimination power showed SVM with polynomial function (100%), followed by radial basis (96.67%) and linear (81.82%).", title = "Machine learning chemometric model for Raman spectroscopy based honey quality assessment", pages = "5-5", url = "https://hdl.handle.net/21.15107/rcub_agrospace_6549" }
Kolašinac, S. M., Pećinar, I.,& Dajić Stevanović, Z. P.. (2021). Machine learning chemometric model for Raman spectroscopy based honey quality assessment. , 5-5. https://hdl.handle.net/21.15107/rcub_agrospace_6549
Kolašinac SM, Pećinar I, Dajić Stevanović ZP. Machine learning chemometric model for Raman spectroscopy based honey quality assessment. 2021;:5-5. https://hdl.handle.net/21.15107/rcub_agrospace_6549 .
Kolašinac, Stefan M., Pećinar, Ilinka, Dajić Stevanović, Zora P., "Machine learning chemometric model for Raman spectroscopy based honey quality assessment" (2021):5-5, https://hdl.handle.net/21.15107/rcub_agrospace_6549 .