Using Machine Learning Models to Predict Hydroponically Grown Lettuce Yield.

Area of Science:

• Agricultural Science
• Machine Learning
• Hydroponics

Background:

• Maximizing global food supply is crucial, especially in developing nations.
• Accurate crop yield prediction aids in efficient food production and resource management.
• Hydroponic systems offer controlled environments for optimizing crop cultivation.

Purpose of the Study:

• To investigate the prediction accuracy of four machine learning (ML) models for lettuce yield (fresh weight).
• To evaluate the performance of Support Vector Regressor (SVR), Extreme Gradient Boosting (XGB), Random Forest (RF), and Deep Neural Network (DNN) models.
• To assess the impact of different hydroponic systems and magnetic field strengths on yield prediction.

Main Methods:

• Lettuce was cultivated in three hydroponic systems under varying magnetic field strengths.
• Four ML models (SVR, XGB, RF, DNN) were employed for yield prediction.
• Three input variable scenarios were tested, including leaf number, water consumption, dry weight, stem length, and stem diameter.

Main Results:

• The XGB model with all input variables achieved the lowest Root Mean Square Error (RMSE) of 8.88 g.
• Support Vector Regressor (SVR) and Deep Neural Network (DNN) models demonstrated excellent prediction accuracy (Scatter Index < 0.1).
• DNN with fewer input variables (scenario 2) was preferred for its efficiency and accuracy.

Conclusions:

• Deep Neural Network (DNN) models show significant potential for predicting fresh lettuce yield accurately.
• Machine learning offers a rapid tool for decision-makers to manage crop yields effectively in hydroponic systems.
• Optimized ML models can contribute to enhancing global food security through improved crop production.