<![CDATA[The rapid advancement of artificial intelligence has enabled practical, data-driven approaches to agricultural quality assessment. However, many existing methods rely on complex sensor systems that are costly and difficult to deploy in the field. This study proposes a lightweight and interpretable K-Nearest Neighbor (KNN) model for non-destructive evaluation of sugarcane milling feasibility using five easily measurable physical attributes: relative distance ratio, internode length, mean diameter, circumference, and weight per centimeter. Samples with Brix less than 16 are categorized as not feasible for milling, while Brix equal to or greater than 16 are classified as possible. A dataset of 1,889 Bululawang samples collected in Malang, East Java, Indonesia, was evaluated across twenty-two scenarios that varied the train-test split, normalization method, distance metric, and neighborhood size. The optimal configuration, consisting of an 80:20 split, Standard normalization, the Minkowski distance metric, and k=75, achieved an accuracy of 78%. The findings confirm that physical measurements can serve as effective predictors of sugarcane quality and support data-driven inspection and sustainable resource utilization in line with SDGs 2, 9, and 12.]]>
Copyrights © 2025
