Development of a Machine Learning Model for Automated Palm Fruit Ripeness Classification

Abstract

Oil palm is one of Indonesia’s most important plantation commodities, yet determining the optimal harvesting time remains a challenge because ripeness assessment is often conducted manually, leading to subjective and inconsistent results. This study aims to conduct a comparative analysis of machine learning models for oil palm fruit ripeness classification using digital images. The dataset, obtained from Mendeley Data, consists of 254 fruit images categorized into three ripeness levels: ripe, half-ripe, and unripe. The research process involved data preprocessing, RGB (Red, Green, Blue) feature extraction, model training using several machine learning algorithms including Support Vector Machine (SVM), Naive Bayes, k-Nearest Neighbors (KNN), Random Forest, Extreme Gradient Boosting (XGBoost), Multi-Layer Perceptron (MLP), and Ensemble Learning, followed by model evaluation using accuracy, precision, recall, and F1-score. The findings show that the SVM model achieved the best performance with an accuracy of 79.41%, followed by MLP with 78.43%. These results indicate that SVM and MLP are capable of effectively distinguishing ripeness levels based on color characteristics. The developed model can support the digitalization of agriculture by enabling more objective, efficient, and consistent fruit classification, thus improving harvest accuracy and productivity. This research demonstrates the potential of machine learning as a reliable tool for automated oil palm ripeness classification in precision agriculture.