Articles
Supervised learning classifiers for differentiating pomegranate cultivars using their multidimensional quality traits
Article number
1464_2
Pages
7 – 12
Language
English
Abstract
Pomegranate cultivars exhibit significant variability in physical, biochemical, textural, phytochemical, and antioxidant properties, which affects their suitability for both fresh market and industrial applications.
Hence, classifying these cultivars is crucial for commercial use, optimizing production, postharvest management, and processing.
Traditional statistical methods are used to explore the variability in fruit quality attributes.
However, machine learning provides a more robust, data-driven approach for distinguishing cultivars.
Therefore, this study aimed to classify eight commercially grown pomegranate cultivars cultivated in South Africa (‘Arakta’, ‘Bhagwa’, ‘Ruby’, ‘Acco’, ‘Ganesh’, ‘Herskowitz’, ‘Molla de Elche’ and ‘Wonderful’) using their multidimensional quality attributes as input data for the machine learning models.
Three supervised learning models, including K-nearest neighbours (KNN), support vector machine (SVM), and multi-layer perceptron (MLP), were trained and evaluated to identify the best-performing model(s) for classifying cultivars.
Our findings demonstrated that all models achieved high classification performance with precision, recall, and F1-score metrics above 0.90 in classifying eight commercially grown pomegranate cultivars cultivated in South Africa using their multidimensional quality attributes as features.
Overall, this study confirms that machine learning algorithms, particularly KNN and SVM (with an accuracy of 99.77% for both), are reliable classifiers for effectively differentiating pomegranate cultivars.
Therefore, while ‘Arakta’ and ‘Ruby’ still require further exploration to achieve accurate distinction, this study highlights the potential of machine learning-driven approaches as robust and efficient tools for pomegranate cultivar selection (targeted breeding), postharvest management, processing, and marketing strategies.
Hence, classifying these cultivars is crucial for commercial use, optimizing production, postharvest management, and processing.
Traditional statistical methods are used to explore the variability in fruit quality attributes.
However, machine learning provides a more robust, data-driven approach for distinguishing cultivars.
Therefore, this study aimed to classify eight commercially grown pomegranate cultivars cultivated in South Africa (‘Arakta’, ‘Bhagwa’, ‘Ruby’, ‘Acco’, ‘Ganesh’, ‘Herskowitz’, ‘Molla de Elche’ and ‘Wonderful’) using their multidimensional quality attributes as input data for the machine learning models.
Three supervised learning models, including K-nearest neighbours (KNN), support vector machine (SVM), and multi-layer perceptron (MLP), were trained and evaluated to identify the best-performing model(s) for classifying cultivars.
Our findings demonstrated that all models achieved high classification performance with precision, recall, and F1-score metrics above 0.90 in classifying eight commercially grown pomegranate cultivars cultivated in South Africa using their multidimensional quality attributes as features.
Overall, this study confirms that machine learning algorithms, particularly KNN and SVM (with an accuracy of 99.77% for both), are reliable classifiers for effectively differentiating pomegranate cultivars.
Therefore, while ‘Arakta’ and ‘Ruby’ still require further exploration to achieve accurate distinction, this study highlights the potential of machine learning-driven approaches as robust and efficient tools for pomegranate cultivar selection (targeted breeding), postharvest management, processing, and marketing strategies.
Authors
Y. Silue, T. Fadji, O.A. Fawole
Keywords
classification, K-nearest neighbours, multi-layer perceptron, physicochemical properties, support vector machine, texture
Online Articles (49)