Comparative Analysis of Machine Learning Techniques for Formalin Fish Classification using Image-based GLCM Feature Extraction

Abstract

With the maturing of artificial intelligence and machine learning, significant advances are being made by researchers in the mainstream artificial intelligence field and experts in other fields who are using these methods to achieve their own goals. Fish is a food whose quality is very susceptible to decline in shape, texture, taste, and smell. Deterioration This damage occurs due to the activity of enzymes and microbiology. Some fishermen choose a fraudulent method to keep their fish fresh, namely mixing them with formaldehyde, while formalin is poisonous if it enters the body. Considering all the problems and consequences of consuming formaldehyde, the author tried to apply several machine-learning methods to classify images of formaldehyde fish on tilapia fish objects. The methods used were a combination of Support Vector Machine (SVM), Multilayer Perceptor (MLP), Naïve Bayes, K-Nearest Neighbor (KNN), and the J48 Decision Tree method to determine which method was the most effective in classifying fish images. Based on the research results, the Multilayer Perceptron (MLP) method achieved better results compared to the other methods, with an accuracy of 0.667. Compared with previous research on the classification of images of tilapia fish, a better accuracy value was achieved.