<![CDATA[ABSTRACTTechnological advancements have accelerated the adoption of various machine learning methods to support decision-making processes in the agricultural and horticultural sectors. One of the persistent challenges in fruit quality assessment is the identification of taste levels, which is commonly performed subjectively based on visual observation and individual experience. Such an approach may lead to inconsistencies in product quality evaluation. Therefore, this study aims to apply the K-Cluster Classification Tree (K-CT) method to classify the taste levels of Gerga oranges based on their physical characteristics. The K-CT method is a hybrid approach that integrates the K-Means Clustering algorithm with a Classification Tree to enhance classification performance while maintaining computational efficiency. The research utilized primary data collected through direct observation of 700 Gerga orange samples obtained from farmers and fruit traders in Tanjung Sakti District, Lahat Regency. Each sample was represented by six physical attributes, namely peel color, pore size, thrips presence, fruit shape, peel texture, and fruit diameter, while taste level served as the target variable. The dataset was divided into 80% training data and 20% testing data. The experimental results demonstrated that the K-CT method achieved a classification accuracy of 94.57%, outperforming Classification Tree, Random Forest, and Gradient Boosting models. Furthermore, the proposed method exhibited competitive computational efficiency and successfully identified fruit diameter as the most influential attribute affecting the taste level of Gerga oranges. These findings indicate that the K-CT method has considerable potential to be implemented as an objective, accurate, and efficient decision support system for fruit quality classification.]]>
Copyrights © 2026
