CLASSIFICATION OF GOLDFISH IMAGES USING K-NEAREST NEIGHBOR (KNN) BASED ON COLOR, SHAPE, AND TEXTURE FEATURES

This study aims to develop an image classification system to automatically identify types of Goldfish (Mas Koki) using the K-Nearest Neighbor (KNN) algorithm. The approach combines three primary feature extraction techniques: color features represented in the HSV color space, shape features obtained through Hu moments, and texture features extracted from grayscale image histograms. The combination of these three features is intended to improve classification performance and address the limitations of previous studies, which typically relied on only one or two types of visual features. The dataset used in this study consists of 1,500 goldfish images categorized into three classes: Oranda, Ryukin, and Ranchu. The KNN model was built with a parameter value of k = 5 and evaluated using four different training and testing data split ratios: 90:10, 80:20, 70:30, and 60:40. Experimental results show that the combination of features effectively enhances classification accuracy, achieving the highest accuracy of 93% with the 80:20 split ratio. This research demonstrates that the KNN method, when paired with appropriate feature extraction, can serve as an effective and efficient solution for developing automatic ornamental fish image classification systems.