Application of the K-Nearest Neighbor (K-NN) Algorithm for Detecting Banana Harvest Feasibility

This study focuses on detecting banana harvest feasibility at the green-ripe stage, an area often overlooked in previous studies that focus only on general ripeness. The objective of this research was to develop a system based on the K-Nearest Neighbor (K-NN) algorithm to classify bananas as “Ready for Harvest” or “Not Ready for Harvest” using digital image processing. The system utilizes Hue Saturation Value (HSV) for color analysis and Gray Level Co-occurrence Matrix (GLCM) for texture identification. Unlike other methods, the combination of HSV and GLCM provides richer, complementary features, improving classification accuracy. The study was conducted at a banana plantation in Kwala Bekala Village, Medan Johor District, with 200 banana images taken from five different locations. The K-NN algorithm, with a value of K = 3, was chosen to avoid tie votes and ensure computational efficiency. The system achieved an accuracy of 94%, with precision of 93.5%, recall of 92.8%, and an F1-score of 93%. In beta testing with 33 respondents (18 farmers and 15 non-farmers), the system achieved a user satisfaction rate of 90%. Misclassifications occurred due to factors such as lighting variations and background noise. The study demonstrates the practical benefit of using the K-NN algorithm for determining the optimal harvest time, helping farmers make more accurate decisions, reducing waste, and increasing market competitiveness. This research fills the gap in detecting green-ripe bananas, providing an innovative solution to optimize harvest timing in the agricultural industry.