Scholarlink

<p>This work introduces a machine learning-aided UV–Vis spectroscopic system for adulteration identification and determination of adulteration in fruit juices. Pomegranate, mango, guava, and pineapple juices were spiked with orange juice in 5 percent to 30 percent concentrations, and absorbance spectra from 300–800&#xa0;nm were observed. Before spectral data analysis, the data were pre-processed with baseline correction and normalization so as to filter out noise and enhance the model robustness. Both the classification and regression models were built on 80:20 training/testing split. The models of classification, such as Random Forest, Gradient Boosting, Support Vector Classifier (SVC), and K-Nearest Neighbors (KNN) had shown good performance and overall accuracy and F1-score were greater than 90, which means that these models were able to discriminate well between the pure and adulterated samples of all types of juices. Quantitative prediction provided different degrees of performance by regression models. CatBoost was the best among them with a value of R<sup>2</sup> of 0.93 (train) and 0.7975 (test) and RMSE of 0.035 (train) and 0.0447 (test), and a value of RPD of 3.1, which is good predictive power. Random Forest and Gradient Boosting models were also characterized by high predictive accuracy (R<sup>2</sup> &gt; 0.75) and moderate predictive accuracy in SVR and KNN. Conversely, Elastic Net regression was also relatively poor because spectral data were nonlinear. The findings prove the fact that UV–Vis spectral characteristics can be used to characterize and predict the adulteration of juices with a high level of accuracy and reliability. It was reported that the wavelength range chosen (300–800&#xa0;nm) was able to retain essential chemical changes related to phenolics, carotenoids and anthocyanins. Overall, the suggested solution is a scalable and effective framework of real-time authentication and quality control of fruit juice.</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

Accelerated detection of fruit juice adulteration through UV–vis spectroscopy and data-driven techniques

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Nivedita Vishwakarma,
  • Amit Suryadipta Kar,
  • Abir Chakravorty

摘要

This work introduces a machine learning-aided UV–Vis spectroscopic system for adulteration identification and determination of adulteration in fruit juices. Pomegranate, mango, guava, and pineapple juices were spiked with orange juice in 5 percent to 30 percent concentrations, and absorbance spectra from 300–800 nm were observed. Before spectral data analysis, the data were pre-processed with baseline correction and normalization so as to filter out noise and enhance the model robustness. Both the classification and regression models were built on 80:20 training/testing split. The models of classification, such as Random Forest, Gradient Boosting, Support Vector Classifier (SVC), and K-Nearest Neighbors (KNN) had shown good performance and overall accuracy and F1-score were greater than 90, which means that these models were able to discriminate well between the pure and adulterated samples of all types of juices. Quantitative prediction provided different degrees of performance by regression models. CatBoost was the best among them with a value of R2 of 0.93 (train) and 0.7975 (test) and RMSE of 0.035 (train) and 0.0447 (test), and a value of RPD of 3.1, which is good predictive power. Random Forest and Gradient Boosting models were also characterized by high predictive accuracy (R2 > 0.75) and moderate predictive accuracy in SVR and KNN. Conversely, Elastic Net regression was also relatively poor because spectral data were nonlinear. The findings prove the fact that UV–Vis spectral characteristics can be used to characterize and predict the adulteration of juices with a high level of accuracy and reliability. It was reported that the wavelength range chosen (300–800 nm) was able to retain essential chemical changes related to phenolics, carotenoids and anthocyanins. Overall, the suggested solution is a scalable and effective framework of real-time authentication and quality control of fruit juice.