Explainable AI-Driven Decision Support System for Early Aquatic Disease Detection and Sustainable Fisheries Production

Abstract

Aquatic diseases pose a major risk to the fisheries and aquaculture sector by causing economic losses, decreased productivity, and environmental instability. Conventional disease surveillance approaches (based on manual exam and lab analysis) are not appropriate for proactive prevention and sustainable fisheries practice. Thus, intelligent early-warning systems are vital for correct disease detection and quick decision making in smart fisheries’ environment. In this study, a decision support system for early detection of aquatic diseases and sustainable production of fisheries products using Explainable Artificial Intelligence (XAI) was proposed. The structure combines water quality parameter, fish health indicators with environmental monitoring data through the IoTs sensing system. A predictive model based on XGBoost is used to perform a disease classification, while the influence of few parameters that are considered very important for the disease, such as dissolved oxygen, pH, ammonia concentration, salinity and temperature, is determined by SHapley Additive exPlanations (SHAP) to enhance the transparency of the model. Accuracy, precision, recall, F1-score, specificity, and ROC-AUC are used to evaluate the system performance. Experimental results show a better accuracy in the prediction, interpretability, proactive disease management, decreased mortality of the fish, and improved sustainability in the aquaculture ecosystem.