Comparative analysis of data using machine learning algorithms: A hydroponics system use case

Abstract

This paper makes a comparison of machine learning algorithms for the analysis of four hydroponic datasets. Data have been gathered daily from hydroponic systems to predict the output of the hydroponic systems. This research compares the performance of the federated split Learning, Deep neural network, extreme Gradient Boosting (XGBoost), and Linear regression algorithms on four different hydroponic systems. These algorithms have been used to analyze the datasets of Nutrient Film Technic (NFT), Floating (FL), Aggregate (AG) and Aeroponic (AER) hydroponic systems. The results have indicated the performance of each model for each hydroponic system and how each algorithm have used the various multiple input features to make predictions of the onion bulb diameter and the errors encountered by each model. From the results obtained, it has been observed that the R square score is varied for each hydroponic system. This variation in the result has been also reflected in the Mean absolute errors obtained. This research determines which of the algorithms predict the optimal Onion bulb diameter (mm) using days after transplant (days), Temperature (°C), water consumption (Litres), Number of Leaves (NL), Nitrogen (mg/g), Phosphorus (mg/g), Potassium (mg/g), Calcium (mg/g), Magnesium (mg/g), Sulphur (mg/g), Sodium (mg/g) as independent variables. The results will be a guide in the choice of hydroponic system to adopt for food production based on the climatic parameters of the location, which is one of the numerous contributions of this research.

Keywords

• Federated split learning
• Nutrient Film Technic
• Aggregate
• Aeroponics
• Floating hydroponics
• Mean absolute error