Scholarlink

<p>The pH differential method is commonly used to determine the total monomeric anthocyanin content in fruits. However, this method requires intensive labor, time, and expensive chemicals. The aim of this study was to develop a&#xa0;simpler and faster model as an alternative to the current method for determining the anthocyanin content of black mulberry fruit. Within the scope of the study, the K60 genotype of black mulberry (<i>Morus nigra</i>), located at the Agricultural Research and Application Center of Tokat Gaziosmanpaşa University, was utilized. The fruits were harvested at various ripening stages in July of 2022 and 2023. Initially, external color measurements (L*, a*,&#xa0;b*) were taken, followed by analyses of pH, soluble solids content (SSC), and anthocyanin levels. The obtained data were categorized into two groups: independent variables (L*, a*,&#xa0;b*, pH, SSC) and the dependent variable (anthocyanin). Four different machine learning algorithms were utilized for anthocyanin prediction: Multilayer Perceptron, Support Vector Machine (SVM), k‑Nearest Neighbors (KNN), and Random Forest. The performance of the developed models was evaluated using the following metrics: correlation coefficient&#xa0;(<i>r</i>), mean absolute error (MAE), root mean square error (RMSE), relative absolute error (RAE), and root relative squared error (RRSE). As a&#xa0;result of the study, among the machine learning algorithms used, the Random Forest algorithm yielded the most successful prediction of anthocyanin values. It was followed by the KNN, SVM, and Multilayer Perceptron algorithms, respectively. The performance metrics of the Random Forest algorithm were determined as follows: <i>r</i> = 0.8498, MAE = 52.779, RMSE = 79.0309, RAE = 43.3826%, and RRSE = 52.7319%. As a&#xa0;result, the findings suggest that machine learning-based models, especially the Random Forest algorithm, hold significant potential for future research focusing on the prediction-based determination of anthocyanin content.</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

Modeling Anthocyanin Content in Black Mulberry (Morus nigra) Fruits Using Machine Learning Based on Color and Chemical Properties

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • Osman Nuri Öcalan,
  • Onur Saraçoğlu

摘要

The pH differential method is commonly used to determine the total monomeric anthocyanin content in fruits. However, this method requires intensive labor, time, and expensive chemicals. The aim of this study was to develop a simpler and faster model as an alternative to the current method for determining the anthocyanin content of black mulberry fruit. Within the scope of the study, the K60 genotype of black mulberry (Morus nigra), located at the Agricultural Research and Application Center of Tokat Gaziosmanpaşa University, was utilized. The fruits were harvested at various ripening stages in July of 2022 and 2023. Initially, external color measurements (L*, a*, b*) were taken, followed by analyses of pH, soluble solids content (SSC), and anthocyanin levels. The obtained data were categorized into two groups: independent variables (L*, a*, b*, pH, SSC) and the dependent variable (anthocyanin). Four different machine learning algorithms were utilized for anthocyanin prediction: Multilayer Perceptron, Support Vector Machine (SVM), k‑Nearest Neighbors (KNN), and Random Forest. The performance of the developed models was evaluated using the following metrics: correlation coefficient (r), mean absolute error (MAE), root mean square error (RMSE), relative absolute error (RAE), and root relative squared error (RRSE). As a result of the study, among the machine learning algorithms used, the Random Forest algorithm yielded the most successful prediction of anthocyanin values. It was followed by the KNN, SVM, and Multilayer Perceptron algorithms, respectively. The performance metrics of the Random Forest algorithm were determined as follows: r = 0.8498, MAE = 52.779, RMSE = 79.0309, RAE = 43.3826%, and RRSE = 52.7319%. As a result, the findings suggest that machine learning-based models, especially the Random Forest algorithm, hold significant potential for future research focusing on the prediction-based determination of anthocyanin content.