Machine Learning Based Smart Aquaponics Farming System

Abstract

For many years, researchers have been studying nutrient management in aquaponic systems. Most have concentrated on adequate nutrition control in an aquaponic setup, but there has been relatively little study on commercial scale applications. For plant growth, it is necessary to measure the level of nutrients present in the soil mixture. In our model, the input data was sourced on some interval of time basis from three commercial aquaponic farms. To rank the features in order of relevance, and feature selection, approaches such as the XGBoost classifier and Recursive Feature Elimination with ExtraTreesClassifier were utilised. Based on the plants and fish, ammonium and nitrates were discovered to be the top two nutrient predictors. The historical dataset's median nutrient levels served as the appropriate concentrations to be maintained in the aquaponic solution in which Ashwagandha was growing. Vernier sensors were utilised to measure nutrient values, and actuator systems were created to discharge the necessary nutrient into the ecosystem through a closed loop. A digital system is created which gives information about the fertiliser(s) required for their crops and the data sensed by the sensors is stored in the cloud and analysed, based on which, suggestions for the growth of the suitable crop are made.