Multi-Layer Perceptron Neural Network and Internet of Things for Improving the Realtime Aquatic Ecosystem Quality Monitoring and Analysis

Abstract

This research proposes improving the aquarium environment for ornamental fish farming for small enterprises raising ornamental fish for sale during the COVID-19 Pandemic with an automatic aquarium system capable of forecasting the optimum environment for fish using Multi-Layer Perceptron Neural Network. Since the amount of the collected data was limited, it was also employed to adjust the imbalanced dataset by applying the Synthetic Minority OverSampling Technique to increase forecasting accuracy. Subsequently, the developed system is based on Internet of Things devices in conjunction with sensors for measuring the indicators that affect the aquarium environment, including temperature, turbidity, total dissolved solids, the potential of hydrogen ions, dissolved oxygen, and nitrate ion. Further, the mobile application was developed and collaborated with sensors and devices to facilitate entrepreneurs monitoring and controlling this automatic system. The results showed that the accuracy of the developed water environment forecasting model was 97.31% and gave the highest level of automation efficiency. Therefore, the developed automated aquarium system could be applied to reduce fish mortality and maintain environmental conditions to grow adequately over the fish life span.