Can Machine Learning classifiers be used to regulate nutrients using small training datasets for aquaponic...

Area of Science:

• Agronomy
• Data Science
• Sustainable Agriculture

Background:

• Aquaponics is a growing alternative to traditional agriculture for sustainable food production.
• A significant challenge in aquaponics is the lack of data-driven methods for cultivation.
• Optimizing nutrient control is crucial for maximizing yield in aquaponic systems.

Purpose of the Study:

• To investigate statistical inference methods for nutrient control in aquaponics using small datasets.
• To enhance yield optimization through data-driven approaches in aquaponic systems.

Main Methods:

• Employed Density-Based Synthetic Minority Over-sampling TEchnique (DB-SMOTE) to address imbalanced datasets.
• Utilized ExtraTreesClassifier and Recursive Feature Elimination (RFE) for feature selection.
• Applied Monte-Carlo (MC) sampling and feature engineering for synthetic data generation.
• Evaluated kernel-based classifiers for nutrient control in aquaponic solutions.

Main Results:

• Successfully applied advanced statistical techniques to small aquaponics datasets.
• Demonstrated effective feature selection and synthetic data generation for improved model performance.
• Showcased the potential for optimizing nutrient control to enhance aquaponic yield.

Conclusions:

• Statistical methods, including DB-SMOTE and feature selection, can effectively address data limitations in aquaponics.
• Data-driven approaches are vital for optimizing nutrient management and improving sustainable food production in aquaponics.
• This research provides a foundation for developing more robust and data-informed aquaponic cultivation strategies.