AquaticSafe Sentinel: An Edge-Based IoT-Powered Predictive System for Continuous Dissolved Oxygen Surveillance in Aquaculture

Aquaculture stands as a flourishing industry, serving as a significant financial resource for many. However, it encounters various challenges such as environmental degradation and disease outbreaks. Consequently, maintaining water quality in aquaculture systems becomes imperative. Dissolved oxygen (DO), a key indicator of oxygen availability for fish respiration, emerges as a crucial measure of water quality. In this research, we propose AquaticSafe Sentinel, an innovative IoT-backed predictive system designed for edge-based continuous monitoring of DO and real-time alerting in aquaculture. Our two-phase framework utilizes both edge computing and cloud computing to swiftly and reliably predict DO levels based on environmental and biological characteristics. Phase one involves making ensemble forecasts of DO levels at the cloud layer using a combination of three machine learning approaches, namely Linear Regression, Random Forest, and AdaBoost. Then phase two involves employing the results of Linear Regression at the edge of the network to make real-time decisions about DO levels, and triggering alerts if necessary. Subsequently, the system awaits a more refined conclusion from the combination technique at the cloud layer. We conduct a comparative analysis with various baseline methods, demonstrating that our framework surpasses them in terms of prediction accuracy and computing efficiency. Additionally, we illustrate how our framework effectively minimizes false alerts, achieving a balance between speed and accuracy to deliver timely solutions.