<![CDATA[Fish catch identification is still frequently performed manually, which may lead to recording errors, particularly when fish species have similar shapes, colors, and textures. This study aims to apply the K-Nearest Neighbor (KNN) algorithm to classify fish catch images in Tanjung Balai City. The research object is limited to three fish species: tuna, mackerel, and Spanish mackerel. The dataset consists of 627 images, divided into 534 training data and 93 testing data. The research stages include image preprocessing using resize with padding at 224 x 224 pixels, color feature extraction using HSV color space, texture feature extraction using Gray Level Co-occurrence Matrix (GLCM), KNN parameter optimization using Grid Search Cross Validation, and evaluation using a confusion matrix. The testing results show an accuracy of 61.29%, precision of 68.65%, recall of 61.29%, and F1-score of 62.83%. The tuna class achieved the best performance, while the Spanish mackerel class was most frequently misclassified. These results indicate that KNN can be used as an initial method for digital image-based fish classification.]]>
Copyrights © 2026
