Machine Learning-Driven Greenhouse Management: A Comparative Analysis of Prediction Models

Abstract

Modern agriculture increasingly relies on smart greenhouses to create controlled environments that optimize crop growth. Effective management of these systems requires accurate prediction of key environmental parameters, robust decision-making processes, and a strong focus on enhancing operational efficiency and sustainability. This study presents a comprehensive evaluation of two machine learning models, the Random Forest Regressor (RFR) and the Gradient Boosting Regressor (GBR), to predict greenhouse conditions for tomato cultivation and manage critical devices such as heating, lighting and irrigation. The models were assessed based on root mean squared error (RMSE), R-squared (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>), and other relevant performance metrics.These results demonstrate the robustness and generalization capacity of both models. RFR exhibited slightly better predictive performance for heating and irrigation, while both models achieved near-identical accuracy for lighting. The high precision of both algorithms highlights their potential for real-time greenhouse management.