Comparative analysis of data driven prediction modeling strategies for aquaculture healthcare

Aquaculture is a significantly important part of our environment. It is the process of rearing, breeding and harvesting of aquatic and marine life in controlled aquatic environments such as seas, oceans and rivers. If the health of aquatic environment is not monitored satisfactorily, conditions like algae bloom, depletion of oxygen and toxins secretions might prevail. This not only contaminates the aquatic environment but is also a threat to entire ecosystem. In this study, we have aimed to predict levels of dissolved oxygen in diverse aquaculture environments. We have established a comprehensive comparative analysis of three well-known and up-to-date machine learning algorithms to predict aquaculture health based on dissolved oxygen levels of the water. We have considered factors such as amount of nitrate, temperature, oxygen saturation and PH level to predict dissolved oxygen (DO). Our results show the high achievement of prediction accuracy of 87.88% by Facebook prophet (FBprophet) as compared to 86.94% by Support Vector Regression (SVR) and 81.06% by Extreme Gradient Boosting (XGBoost).