Classification of Fish Species Using the K-Nearest Neighbor (KNN) Method: A Case Study at Bintan Center Market, Tanjungpinang

Abstract

Indonesia has a diversity of fish species, but the identification process is still largely done manually, which is prone to errors and less efficient. This research aims to develop an automatic fish species classification system using the K-Nearest Neighbor (KNN) algorithm. Data were collected directly at Bintan Center Market Tanjungpinang, focusing on four types of fish: Sardinella, Scomber, Barbonymus and Euthynnus. Visual features of the fish such as texture and shape were extracted from digital images as classification features. The KNN algorithm then compares new data with classified data to determine fish species based on the majority of nearest neighbors. The research results show that the KNN method can improve classification accuracy and efficiency in identifying fish species. The feature extraction used in this study includes: GLCM (Gray Level Co-occurrence Matrix) and HOG (Histogram of Oriented Gradients). Preliminary training results show that the system achieved an accuracy of 71.34% on the training dataset, the results suggest that the method is promising for improving classification accuracy and efficiency in identifying fish species.