Robust Fan Actuator Prediction in Smart Greenhouses Using Machine Learning: A Comparative Analysis of Ensemble and Linear Models

The increasing demand for sustainable agriculture has driven the development of smart greenhouses equipped with automated systems for climate control. A critical component of these systems is the fan actuator, which regulates airflow and stabilizes the internal climate. This study explores the use of machine learning models for predicting the activation status of fan actuators based on environmental data collected from a smart greenhouse. We evaluate several machine learning models, including Support Vector Machine (SVM), Random Forest, Gradient Boosting, XGBoost, and Logistic Regression, under real-world conditions simulated by adding noise and label corruption to the dataset. The dataset was augmented and balanced using the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalances. Results indicate that ensemble methods, particularly XGBoost and Random Forest, outperform simpler models in terms of accuracy, precision, recall, and F1 score. XGBoost achieved the highest accuracy at 94.47%, while Random Forest followed closely with 94.29%. The study demonstrates that these models are robust to data imperfections and can be effectively employed for real-time fan actuator control. However, further validation is needed to generalize the findings to different greenhouse environments. The research highlights the potential of machine learning models to improve operational efficiency in smart farming, offering insights into how these technologies can support more sustainable agricultural practices.