International Journal of Science, Engineering and TechnologyAuthors: Dr. Pankaj Malik, Anmol Singh Tomar, Ayush Trivedi, Aniruddha Paliwal, Ansh Goyal
Abstract: Efficient and site-specific fertilizer application is a cornerstone of precision agriculture, aiming to enhance crop yield while minimizing environmental impact. Traditional fertilizer practices often lead to overuse or under-application, resulting in resource inefficiency, soil degradation, and reduced profitability. In this study, we propose a machine learning-based system for optimizing fertilizer application by analyzing key agronomic parameters such as soil nutrients (N, P, K), pH, organic carbon, weather conditions (temperature, rainfall), and crop type. We evaluated several machine learning models, including Random Forest, Artificial Neural Networks, and XGBoost, using the publicly available Soil and Crop Fertilizer Recommendation Dataset. The experimental results show that the XGBoost model achieved the best performance with an accuracy of 93.4%, F1-score of 0.92, and AUC of 0.96 in predicting the optimal fertilizer type and dosage. Field-level simulations further demonstrated a 17% increase in average crop yield and a 23% reduction in fertilizer usage compared to traditional application methods. These findings suggest that machine learning can play a significant role in advancing sustainable agricultural practices by delivering intelligent, data-driven fertilizer recommendations.
DOI: https://doi.org/10.5281/zenodo.15737088
