Scholarlink

<p>Food fraud involving visually misleading adulterants, such as paprika mixed with color-matched substances, poses a growing challenge to food safety authorities. To address this issue, a rapid and non-destructive authentication method was developed. This study aims to combine Visible and Near-Infrared Spectroscopy (Vis/NIRS) with Machine Learning (ML) algorithms to authenticate paprika adulterated with flour colored with red paint, an unexamined adulterant. Spectral data were collected using two spectrometers operating in the 500–1000&#xa0;nm and 1000–1700&#xa0;nm spectral ranges. To optimize model performance and mitigate the overfitting risk, Variable Iterative Space Shrinkage Approach (VISSA) and interval VISSA (iVISSA) were employed as Feature Selection (FS) methods. In addition, Dimensionality Reduction (DR) techniques, including Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE), were evaluated. Three ML algorithms were compared, namely Support Vector Regression (SVR), Artificial Neural Networks (ANN), and Gaussian Process Regression (GPR). Among all configurations tested, VISSA-PCA-ANN achieved the best results in the NIR spectral range, with a prediction determination coefficient R<sup>2</sup><sub>p</sub> = 0.99 and a Root Mean Square Error of prediction RMSEP = 0.85%. These findings demonstrate the effectiveness of the proposed approach for accurate paprika adulteration quantification, offering a practical solution for food quality control and fraud prevention.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantification of paprika adulteration using visible/near-infrared spectroscopy and machine learning

  • 期刊论文

  • 发表日期：

    2025年10月27日

查看全文
  • Youssef Tmimi,
  • Mounim Chikri,
  • Loubna Srata,
  • Sofia Farres,
  • Fouad Fethi

摘要

Food fraud involving visually misleading adulterants, such as paprika mixed with color-matched substances, poses a growing challenge to food safety authorities. To address this issue, a rapid and non-destructive authentication method was developed. This study aims to combine Visible and Near-Infrared Spectroscopy (Vis/NIRS) with Machine Learning (ML) algorithms to authenticate paprika adulterated with flour colored with red paint, an unexamined adulterant. Spectral data were collected using two spectrometers operating in the 500–1000 nm and 1000–1700 nm spectral ranges. To optimize model performance and mitigate the overfitting risk, Variable Iterative Space Shrinkage Approach (VISSA) and interval VISSA (iVISSA) were employed as Feature Selection (FS) methods. In addition, Dimensionality Reduction (DR) techniques, including Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE), were evaluated. Three ML algorithms were compared, namely Support Vector Regression (SVR), Artificial Neural Networks (ANN), and Gaussian Process Regression (GPR). Among all configurations tested, VISSA-PCA-ANN achieved the best results in the NIR spectral range, with a prediction determination coefficient R2p = 0.99 and a Root Mean Square Error of prediction RMSEP = 0.85%. These findings demonstrate the effectiveness of the proposed approach for accurate paprika adulteration quantification, offering a practical solution for food quality control and fraud prevention.