Improved Eliminate Particle Swarm Optimization on Support Vector Machine for Freshwater Fish Classification

Abstract

Freshwater fish is a type of fish for consumption to substitute saltwater fish in areas far from the sea. Freshwater fish production is not as much as saltwater fish because the selection of freshwater fish is not suitable in the cultivation process. Therefore, a classification method to determine the type of freshwater fish that is good for aquaculture is proposed, one of the classification methods that can be used is the Support Vector Machine (SVM). SVM is used to process the classification of freshwater fish that have the potential to be cultivated. SVM has weaknesses in the classification process, so SVM requires an optimization method that can overcome it's weaknesses, one optimization method that can be used is Improved Eliminate Particle Swarm Optimization (IEPSO). IEPSO is used to optimize features and parameters in the SVM method. The data used in the research came from the Fish and Livestock Breeding Center of Nganjuk Regency with a percentage of 90% of training data and 10% of test data. The results of the study using SVM-IEPSO using feature selection and parameter optimization obtained 88% accuracy in the classification process to determine the type of freshwater fish. The classification results will be more optimal if there are more types of freshwater fish species used in the training process, because this study only uses data on freshwater fish species that are only found in Livestock Breeding Center of Nganjuk Regency.