AI-Powered Optimization in Vertical Farming: Predicting Crop Yield Using Environmental Parameters

Increasing demand for sustainable agricultural practices has driven the adoption of controlled-environment agriculture, such as vertical farming. In this study, we propose an AI-powered approach to optimize crop yield prediction using environmental parameters within vertical farming systems. A publicly available dataset from Kaggle was used, containing over 36,000 records of environmental conditions and corresponding crop yields. After preprocessing and feature engineering, several regression models were evaluated, including Linear Regression, Random Forest, Gradient Boosting, and XGBoost. Among them, XGBoost achieved the best performance with an R2 score of 0.9776 and RMSE of 3.99. To ensure model interpretability, SHAP (SHapley Additive exPlanations) was applied, revealing key environmental features that influence yield outcomes. This work demonstrates the effectiveness of machine learning models-especially XGBoost-when combined with explainable AI techniques for precision agriculture and offers a reliable framework for real-time yield forecasting in vertical farming systems. Our novel contribution lies in demonstrating the synergistic power of advanced machine learning (XGBoost) and Explainable AI (SHAP) to not only achieve unprecedented predictive accuracy in vertical farming but, more critically, to demystify the ‘black box’ and provide actionable, interpretable insights for real-time environmental control, a gap largely unaddressed in existing literature.