Abstract—Precision agriculture increasingly requires reliable, data-driven tools to assist farmers in selecting crops suited to local soil and climate conditions, thereby enhancing productivity and sustainability. This study presents a comprehensive comparative analysis of six advanced machine learning models for smart crop recommendation, leveraging soil and environmental data to identify optimal crops under specific growing conditions. The research addresses a multi-class classification problem using the Kaggle Crop Recommendation Dataset, which comprises 2200 observations across 22 crop types and seven key environmental parameters: nitrogen, phosphorus, potassium, temperature, humidity, pH, and rainfall. Results demonstrate exceptional performance from gradient boosting methods, with Extreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM) achieving 99.77% accuracy, followed by Random Forest at 96.36%. Deep learning approaches (Gated Recurrent Unit (GRU): 91.82%, Long Short-Term Memory (LSTM): 90.45%) and a hybrid Random Forest–LSTM model (88.41%) performed competitively but less effectively. Feature-importance analysis proved that soil nutrients (NPK) were dominant predictors, contributing approximately 60% to the overall importance in ensemble models. Statistical testing confirmed significant performance differences (p < 0.001), while computational analysis favored LightGBM, with the shortest training time (1.8 min) and lowest memory usage (198 MB). These findings provide evidence-based guidance for agricultural technology developers and practical insights for deploying robust crop recommendation systems across diverse farming contexts. Ethical considerations regarding data bias and equitable access to Artificial Intelligence (AI) tools are discussed, emphasizing inclusive and sustainable development practices.
 
Keywords—crop recommendation, precision agriculture, machine learning, ensemble learning, Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), ethical AI


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).