Comparative Study on Prediction Algorithms for Paralichthys olivaceus Diseases Using External and Internal Symptoms

Fish that live in fish farms spread very quickly when they get sick, so if they do not prevent it in advance, they will cause significant losses to fish farms. Currently, research is actively being conducted to detect and predict external and internal disease symptoms of fish through various deep learning technologies. However, even for a specific disease, the symptoms that are expressed for each disease are not clear, and symptoms are frequently expressed across multiple diseases, so the disease classification performance based on symptoms is still low. In this study, we use deep learning techniques to find external and internal disease symptoms in fish, propose various methodologies to predict disease in fish, and select the best disease prediction algorithm through performance comparison. We propose a total of three methodologies as disease prediction algorithms: a method based on a disease symptom table analyzed by a fish disease expert, a disease prediction method based on data statistics, an ensemble model, and a prediction method based on deep learning model training. Experiments show that disease symptom tables and data statistics-based prediction methods performed 48.45% and 50.91%, respectively, while disease prediction methods using ensemble model and deep learning model training performed 70.10%, showing the highest prediction accuracy.