AI-Driven Smart Agriculture: An Integrated Approach for Soil Analysis, Irrigation, and Crop-Fertilizer Recommendations

Modern agriculture faces critical challenges, including climate change induced water shortage, inefficient resource utilization, and reducing soil health which reduce sustainability and reduce crop yields. Traditional farming methods, dependent on manual observation and generalized practices, often lack the precision needed to address these growing issues. To meet these demands, this study presents an artificial intelligence (AI) powered smart agriculture system that integrates deep learning (DL), Internet of Things (IoT), and explainable artificial intelligence (XAI) to enable intelligent, data-driven farming. This study addresses key challenges such as ineffective soil management, poor irrigation practices, and the absence of personalized crop and fertilizer recommendations. IoT-based sensors provide real-time soil analysis, measuring critical parameters such as nutrient levels, moisture, and temperature. Advanced DL models were employed for various tasks in the AI-Driven Smart Agriculture System. The Transformer-based Tabular Learning (TTL) model for irrigation advice gives an accuracy of 99.13%. For soil analysis, the TabNet regressor achieved a root mean square error (RMSE) of 1.51 and an R2 score of 98.7%, while XAI techniques were used to explain and interpret its predictions. The Sparse Weighted Fusion Transformer (SwiFT) model for crop recommendation gives an accuracy of 98.75%. For fertilizer recommendation, the TabNet classifier achieved an accuracy of 99.3%, with XAI techniques applied to explain the reasoning behind its predictions. These results demonstrate that the system delivers precise and actionable insights. A weather Application Programming Interface (API) is integrated to support timely irrigation decisions through accurate forecasting. A chatbot interface is also added which enhances system usability, accessibility and user-friendliness for farmers. The proposed model improves crop productivity, optimizes resource use, and supports adaptation to environmental changes. By leveraging AI and IoT technologies, the proposed model gives a solution which helps in sustainable agriculture, environmental resilience, and global food security, thus benefiting stakeholders ranging from individual farmers to large-scale agricultural business. Future work will explore real-time adaptive control mechanisms and expand model generalization across diverse geographical region.