Classification of Unifloral Honeys from SARDINIA (Italy) by ATR-FTIR Spectroscopy and Random Forest

Tarazona

2021

Foods

Unifloral honeys are highly demanded by honey consumers, especially in Europe. To ensure that a honey belongs to a very appreciated botanical class, the classical methodology is palynological analysis to identify and count pollen grains. Highly trained personnel are needed to perform this task, which complicates the characterization of honey botanical origins. Organoleptic assessment of honey by expert personnel helps to confirm such classification. In this study, the ability of different machine learning (ML) algorithms to correctly classify seven types of Spanish honeys of single botanical origins (rosemary, citrus, lavender, sunflower, eucalyptus, heather and forest honeydew) was investigated comparatively. The botanical origin of the samples was ascertained by pollen analysis complemented with organoleptic assessment. Physicochemical parameters such as electrical conductivity, pH, water content, carbohydrates and color of unifloral honeys were used to build the dataset. The following ML algorithms were tested: penalized discriminant analysis (PDA), shrinkage discriminant analysis (SDA), high-dimensional discriminant analysis (HDDA), nearest shrunken centroids (PAM), partial least squares (PLS), C5.0 tree, extremely randomized trees (ET), weighted k-nearest neighbors (KKNN), artificial neural networks (ANN), random forest (RF), support vector machine (SVM) with linear and radial kernels and extreme gradient boosting trees (XGBoost). The ML models were optimized by repeated 10-fold cross-validation primarily on the basis of log loss or accuracy metrics, and their performance was compared on a test set in order to select the best predicting model. Built models using PDA produced the best results in terms of overall accuracy on the test set. ANN, ET, RF and XGBoost models also provided good results, while SVM proved to be the worst.

Section: Discussionmentioning

confidence: 99%

Comparative Study of Several Machine Learning Algorithms for Classification of Unifloral Honeys

Tarazona

2021

Foods

“…To ensure fair trading of such high valued products as honey, an authentication system needs to be established to detect fraud in the honey supply chain. Most reported works on honey authentication mainly focus on the floral type (botanical origin) and geographical origins, and less frequently on entomological origins [10][11][12][13][14]. Few reported works have been reported on the honey authentication based on entomological origins [15,16].…”

Section: Introductionmentioning

confidence: 99%

The Use of UV Spectroscopy and SIMCA for the Authentication of Indonesian Honeys According to Botanical, Entomological and Geographical Origins

Suhandy

Yulia

2021

Molecules

As a functional food, honey is a food product that is exposed to the risk of food fraud. To mitigate this, the establishment of an authentication system for honey is very important in order to protect both producers and consumers from possible economic losses. This research presents a simple analytical method for the authentication and classification of Indonesian honeys according to their botanical, entomological, and geographical origins using ultraviolet (UV) spectroscopy and SIMCA (soft independent modeling of class analogy). The spectral data of a total of 1040 samples, representing six types of Indonesian honey of different botanical, entomological, and geographical origins, were acquired using a benchtop UV-visible spectrometer (190–400 nm). Three different pre-processing algorithms were simultaneously evaluated; namely an 11-point moving average smoothing, mean normalization, and Savitzky–Golay first derivative with 11 points and second-order polynomial fitting (ordo 2), in order to improve the original spectral data. Chemometrics methods, including exploratory analysis of PCA and SIMCA classification method, was used to classify the honey samples. A clear separation of the six different Indonesian honeys, based on botanical, entomological, and geographical origins, was obtained using PCA calculated from pre-processed spectra from 250–400 nm. The SIMCA classification method provided satisfactory results in classifying honey samples according to their botanical, entomological, and geographical origins and achieved 100% accuracy, sensitivity, and specificity. Several wavelengths were identified (266, 270, 280, 290, 300, 335, and 360 nm) as the most sensitive for discriminating between the different Indonesian honey samples.

“…Devi et al [17] also reported a satisfactory classification among multifloral and monofloral honey when ATR-FTIR was coupled with PCA. Finally, Ciulu et al [48] applied a random forest algorithm to the classification of 80 samples of asphodel, eucalyptus, thistle, and strawberry tree honey, thereby achieving an average accuracy of 87% in a cross-validation context.…”

Section: Stepwise Lda Based On Atr-ftir Spectramentioning

confidence: 99%

SPME-GC-MS and FTIR-ATR Spectroscopic Study as a Tool for Unifloral Common Greek Honeys’ Botanical Origin Identification

Xagoraris

Revelou

Dedegkika³

et al. 2021

Applied Sciences

Among the variants of Greek honey, the most commonly available are pine, fir, thyme, and citrus honey. Samples of the above kinds of honey, identified according to European and Greek legislation, were studied using gas chromatography coupled with mass spectrometry (GC-MS) and the attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopic techniques. Two chemometric models were developed based on statistically significant volatile compounds (octane; 2-phenylacetaldehyde; 1-nonanol; methyl 2-hydroxybenzoate; 2-(4-methylcyclohex-3-en-1-yl); nonanoic acid) and the 1390–945 and 847–803 cm−1 spectral regions, mainly vibrations of fructose and glucose, combined with the stepwise linear discriminant analysis (stepwise LDA) statistical technique. In total, 85.5% of standard samples, and 82.3% through internal validation and 88.5% through external validation, were identified correctly using the GC-MS-stepwise-LDA chemometric model. The corresponding results for the ATR-FTIR-stepwise-LDA chemometric model were 93.5%, 82.5%, and 84.6%. The double validation (internal, external) enhances the robustness of the proposed chemometric models. The developed models are considered statistically equivalent, but FTIR spectroscopy is simple, rapid, and more economical.