Scholarlink

This study presents optimization and machine learning (ML) methods for predicting pigment concentrations in Extra Virgin Olive Oil (EVOO) using UV-Vis spectroscopy data and compares them to the standard deconvolution approach. We tested three EVOO samples: standard, fresh, and Monocultivar Frantoio. All methods demonstrated great spectrum reconstruction accuracy of R2 > 0.99. The optimization technique, which employed Non-Negative Least Squares (NNLS) and Limited-memory Broyden-Fletcher-Goldfarb-Shanno with Bounds (L-BFGS-B) algorithms, consistently yielded non-negative pigment concentration predictions. Both new methods predicted higher total pigment concentrations than the traditional method, with differences ranging from 2.124 mg/kg (Fresh EVOO) to 7.329 mg/kg (Monocultivar Frantoio). Our approach is a major step forward in quick, non-destructive EVOO analysis, even though there are still difficulties in quantifying trace pigments below specific thresholds. The EVOO quality assessment, authentication, and industrial quality control procedures are significantly impacted by these advancements in pigment quantification accuracy and dependability, which also present opportunities for automated, high-throughput screening applications.

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

Machine Learning Techniques for Predicting Olive Oil Pigment Concentrations

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Srikrishna S. Kashyap,
  • B. K. Vinay,
  • T. R. Suranjan,
  • M. A. Sriraj

摘要

This study presents optimization and machine learning (ML) methods for predicting pigment concentrations in Extra Virgin Olive Oil (EVOO) using UV-Vis spectroscopy data and compares them to the standard deconvolution approach. We tested three EVOO samples: standard, fresh, and Monocultivar Frantoio. All methods demonstrated great spectrum reconstruction accuracy of R2 > 0.99. The optimization technique, which employed Non-Negative Least Squares (NNLS) and Limited-memory Broyden-Fletcher-Goldfarb-Shanno with Bounds (L-BFGS-B) algorithms, consistently yielded non-negative pigment concentration predictions. Both new methods predicted higher total pigment concentrations than the traditional method, with differences ranging from 2.124 mg/kg (Fresh EVOO) to 7.329 mg/kg (Monocultivar Frantoio). Our approach is a major step forward in quick, non-destructive EVOO analysis, even though there are still difficulties in quantifying trace pigments below specific thresholds. The EVOO quality assessment, authentication, and industrial quality control procedures are significantly impacted by these advancements in pigment quantification accuracy and dependability, which also present opportunities for automated, high-throughput screening applications.