Scholarlink

<p>This study uses portable spectroscopy, spectral preprocessing, principal component analysis (PCA), and machine learning (ML) to detect and quantify formaldehyde adulteration in buffalo milk. A calibration dataset was created by spiking milk samples with varying concentrations of formaldehyde (0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0, 10.0, 20.0, 30.0, 40.0, and 50.0%) to simulate varying levels of adulteration. Spectral data were acquired in the near-infrared (NIR) range (900–1700&#xa0;nm) using a portable spectrophotometer. Model performance was assessed using two concentration ranges: a low-level adulteration subset (0–10%) to mimic realistic adulteration scenarios, and the entire dataset (0–50%) to evaluate performance across a broader range. The regression model had coefficient of determination (R<sup>2</sup>) values of 0.922–0.998, RMSE of 0.745 − 0.724, and RPD of 3.582–20.490 for 0–10% and 0–50% ranges, respectively, suggesting good to exceptional predictive performance. In classification, the 10-fold cross-validated Matthews correlation coefficient (MCC) ranged from 0.845 ± 0.050 (0–10%) to 0.922 ± 0.019 (0–50%), indicating robust, stable performance. These findings suggest the use of portable NIR spectroscopy in conjunction with machine learning for quick, non-destructive identification of formaldehyde in milk.</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

Portable NIR spectroscopy and machine learning for quantifying formaldehyde adulteration in buffalo milk

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Delinka Genoveva Rosa,
  • Vishant V. Malik,
  • Aniket Gaonkar,
  • Lalchand B. Patle,
  • Jivan S. Parab,
  • Madhusudan G. Lanjewar

摘要

This study uses portable spectroscopy, spectral preprocessing, principal component analysis (PCA), and machine learning (ML) to detect and quantify formaldehyde adulteration in buffalo milk. A calibration dataset was created by spiking milk samples with varying concentrations of formaldehyde (0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0, 10.0, 20.0, 30.0, 40.0, and 50.0%) to simulate varying levels of adulteration. Spectral data were acquired in the near-infrared (NIR) range (900–1700 nm) using a portable spectrophotometer. Model performance was assessed using two concentration ranges: a low-level adulteration subset (0–10%) to mimic realistic adulteration scenarios, and the entire dataset (0–50%) to evaluate performance across a broader range. The regression model had coefficient of determination (R2) values of 0.922–0.998, RMSE of 0.745 − 0.724, and RPD of 3.582–20.490 for 0–10% and 0–50% ranges, respectively, suggesting good to exceptional predictive performance. In classification, the 10-fold cross-validated Matthews correlation coefficient (MCC) ranged from 0.845 ± 0.050 (0–10%) to 0.922 ± 0.019 (0–50%), indicating robust, stable performance. These findings suggest the use of portable NIR spectroscopy in conjunction with machine learning for quick, non-destructive identification of formaldehyde in milk.