Under a realistic cellular metabolic economic model of Nile tilapia, this work investigates Q-learning fish growth trajectory analysis. we offer two Q-learning data mining algorithms the best management policy using simulated data of optimum growth projections starting at the early growth stage and ending at appropriate market share. The early or primitive Q-learning technique learns optimum feed supplying regulatory regime for fish cultivation rate for the fished present the cages, whereas other one continuously adjusts appropriate nutrition controls the law for agricultural fish rate of increase in tanks together within ideal temperature pattern. The cumulative percentage weight of fishes through both vessels on land and afloat cages was tracked with relative dynamic model errors of 1.7% and 6.6%, respectively, using Q-learning control rules. Furthermore, contrasted to drifting cages where its water temp is controlled at 29.7 ℃, the eating and heat management policies minimise the proportional rate of feeding along with wastage of food in by 11%. Data, algorithms, and performance of DL approaches used in adaptive aquaculture farms are also examined. In a nutshell, our goal is to give scholars and clinicians a greater grasp of the latest advancements of deep learning in aquaculture, which will help them develop smart aquacultural applications. With improved machine - learning technologies, a video trying to track biological early warning has made significant progress. Artificial intelligence has helped researchers better understand behavioural reactions to pharmacological and environmental stress. We discuss pioneer efforts in exact monitoring of a gathering of individuals in 2d/3d space in this paper, which introduces the fundamentals of activity recognition. Video tracking's professional and operational difficulties are discussed. The toxicological study based on fish behavioral data is then summarized.