A Transfer Learning-Based Ensemble Approach for Crop Recommendation based on Environmental conditions

Abstract— Accurate crop selection plays a crucial role in maximizing agricultural productivity and ensuring food security, especially in regions with diverse climatic and soil conditions. Traditional farming practices often rely on experience-based decisions, which may not adapt well to changing environmental factors. This research proposes a robust crop recommendation system based on transfer learning and ensemble approach that use environmental and soil parameters such as nitrogen (N), phosphorus (P), potassium (K), temperature, humidity, pH, and rainfall. Classical classifiers including Decision Tree, Random Forest, Naïve Bayes, and Support Vector Machine were used to establish baseline performance. Although some models achieved high accuracy (e.g., Decision Tree and Naïve Bayes at 100%), they exhibited poor precision and F1-scores, indicating overfitting and unreliable generalization. To overcome this, a transfer learning-based ensemble model was developed, which achieved significantly improved and balanced results, with 99% accuracy, precision, recall, and F1-score. This approach demonstrates strong generalization capabilities, particularly in data-scarce regions, by effectively transferring learned patterns from rich datasets to underrepresented domains. The proposed model offers a scalable, data-driven solution for intelligent crop planningand can be extended for real-time deployment to support precision agriculture. Keywords— Crop Recommendation, Machine Learning, Ensemble Model, Environmental Parameters, Transferlearning