XAI-Powered Smart Agriculture Framework for Enhancing Food Productivity and Sustainability

A vital component of maintaining the world’s expanding populace is farming. In agriculture field, factors such as soil quality, weather patterns, and crop yields are essential components of usual possessions that affect farming manufacture. Despite advancements, prevailing smart systems quiet struggle with handling big amounts in prediction claims, often facing difficulties in balancing prediction accuracy and learning efficiency. To ensure sustainable food production, integrating advanced machineries such as machine learning and artificial intelligence in agriculture is essential. This study proposes an explainable AI (XAI)-based smart agriculture system to provide holistic recommendation for precision farming aimed at improving productivity while reducing environmental impact. We compiled a comprehensive weather, soil, and crop dataset from official and verified sources in India. From this dataset, we extracted and optimized features using pre-trained architectures and enhanced barnacles mating optimization (EBMO) algorithm, addressing the high-measure mentality and computational complexity issues often encountered in agricultural data analysis. The selected features were analyzed to provide holistic recommendations for precision farming using baseline ML models such as support vector machine, random forest, neural network, and decision tree. Additionally, we integrate the XAI framework with an interpretable recommendation/s for optimizing the agricultural practices. The study developed an XAI-based smart agriculture system that provides holistic recommendations for precision farming to boost productivity. By using a comprehensive dataset and optimizing features with the EBMO algorithm, the research achieved high accuracy in crop yield predictions, particularly with the XLNet+SVM model, which outperformed existing models across various crops. The integration of SHapley Additive explanation (SHAP) and Local Interpretable Model-agnostic Explanation (LIME) further enabled interpretable AI-driven insights, enhancing transparency in decision-making. The results demonstrated significant improvements in prediction accuracy, resource management, and sustainability, offering valuable contributions to global food security through smart agriculture. The ability to explain these AI decisions further supports the adoption of AI technologies in agriculture, fostering resilient agricultural systems that are essential for feeding the world’s growing population.