Three-dimensional fluorescence combined with alternating trilinear decomposition and random forest algorithm for the rapid prediction of species, geographical origin and main components of Glycyrrhizae Radix et Rhizoma (Gancao)

Glycyrrhizae Radix et Rhizoma (Gancao) is a functional food whose quality varies significantly between distinct geographical sources owing to the influence of genetics and the geographical environment. This study employed three-dimensional fluorescence coupled with alternating trilinear decomposition (ATLD) and random forest (RF) algorithms to rapidly predict Gancao species, geographical origins, and primary constituents. Seven fluorescent components were resolved from the three-dimensional fluorescence of the ATLD for subsequent analysis. Results indicated that the RF model distinguished Gancao from various species and origins better than other algorithms, achieving an accuracy of 94.4 % and 88.9 %, respectively. Furthermore, the RF regressor algorithm was used to predict the concentrations of liquiritin and glycyrrhizic acid in Gancao, with 96.4 % and 95.6 % prediction accuracies compared to HPLC, respectively. This approach offers a novel means of objectively evaluating the origin of food and holds substantial promise for food quality assessment.

Development of machine learning models using multi-source data for geographical traceability and content prediction of Eucommia ulmoides leaves

Rapid and scientific quality evaluation is a hot topic in the research of food and medicinal plants. With the increasing popularity of derivative products from Eucommia ulmoides leaves, quality and safety have attracted public attention. The present study utilized multi-source data and traditional machine learning to conduct geographical traceability and content prediction research on Eucommia ulmoides leaves. Explored the impact of different preprocessing methods and low-level data fusion strategy on the performance of classification and regression models. The classification analysis results indicated that the partial least squares discriminant analysis (PLS-DA) established by low-level fusion of two infrared spectroscopy techniques based on first derivative (FD) preprocessing was most suitable for geographical traceability of Eucommia ulmoides leaves, with an accuracy rate of up to 100 %. Through regression analysis, it was found that the preprocessing methods and data blocks applicable to the four chemical components were inconsistent. The optimal partial least squares regression (PLSR) model based on aucubin (AU), geniposidic acid (GPA), and chlorogenic acid (CA) had a residual predictive deviation (RPD) value higher than 2.0, achieving satisfactory predictive performance. However, the PLSR model based on quercetin (QU) had poor performance (RPD = 1.541) and needed further improvement. Overall, the present study proposed a strategy that can effectively evaluate the quality of Eucommia ulmoides leaves, while also providing new ideas for the quality evaluation of food and medicinal plants.

Detection of the 5-hydroxymethylfurfural content in roasted coffee using machine learning based on near-infrared spectroscopy

Since 5-hydroxymethylfurfural (5-HMF) is carcinogenic to humans, its detection in foods is essential. This study performed near-infrared (NIR) spectroscopy (11998-4000 cm^-1) to determine the 5-HMF content in roasted coffee. The random forest (RF) was used to extract important wavenumbers, after which three machine learning models (ordinary least square (OLS), support vector machine (SVM), and RF) were established for the prediction. RF obtained the best prediction results (R_c² = 0.98 and R_p² = 0.92) compared with OLS and SVM and effectively extracted the important wavenumbers (11667 cm^-1, 11666 cm^-1, 10905 cm^-1, 7096 cm^-1, 7095 cm^-1, 7094 cm^-1, 7093 cm^-1, 7092 cm^-1, 5054 cm^-1, 5026 cm^-1, 5025 cm^-1, and 5024 cm^-1). The results demonstrated that machine learning models based on NIR spectroscopy could provide a non-destructive approach for determining 5-HMF content in roasted coffee.

Identification of different bran-fried Atractylodis Rhizoma and prediction of atractylodin content based on multivariate data mining combined with intelligent color recognition and near-infrared spectroscopy

Unclear established standard of bran-fried Atractylodis Rhizoma (BFAR), a commonly used drug in Traditional Chinese Medicine (TCM), compromised its clinical efficacy. In this study, we explored the correlation between color and near-infrared spectroscopy (NIR) feature with content of atractylodin, then established a rapid recognition model for the optimal degree of processing for BFAR preparation. The results of the Pearson analysis indicated that the color values were significantly and positively correlated with atractylodin content. The back propagation artificial neural network algorithm and cluster analysis revealed the color of different BFAR could be accurately divided into three categories; subsequently, the color range for the optimal degrees of stir-frying was established as follows: R[red value (105.79-127.25)], G[green value(75.84-89.64)], B[blue value(33.33-42.73)], L[Lightness (81.26-95.09)].Using NIR, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and cluster analysis, three types of BFAR were accurately identified. The prediction model of atractylodin content was established using partial least squares regression analysis. The R² of the validation set was 0.9717 and the root mean square error was 0.026. In the color judgment model, the processing degree of 8 batches of BFAR from the market is inferior. According to the NIR judgment model, the processing degree of all samples from the market is inferior. In conclusion, the best fire degree of BFAR can be identified quickly and accurately based on our established model. It is a potential method for quality evaluation of Chinese Materia Medica processing.