Fertilizer and Crop Recommendation using Multi-Model Stacked Learning with Explainable AI for Sustainable Farming

Crop yields are crucial for food security, sustainable agriculture, and economic growth. Fertilizers supply plants with the nutrients such as nitrogen, phosphorus, and potassium required for optimal growth. Farmers, however, lack knowledge and resources to identify suitable soil properties and climatic conditions and consequently make incorrect crop and fertilizer choices. Misuse or excess use of fertilizers degrade the land, damage yield and the environment. This research addresses these problems through an intelligent data-driven solution. This work proposes a Crop and Fertilizer Recommendation System based on Ensemble Learning and Explainable AI. The model integrates Random Forest, Extreme Gradient Boosting, and Long Short-Term Memory as base learners and Logistic Regression as a meta-model in a stacked ensemble architecture. This architecture enhances predictive performance, model generalization, and stability. Models were evaluated using accuracy, precision, recall, and F1-score as performance metrics. To obtain interpretability and trust, the system uses Shapley Additive Explanations, which offer insights into the contribution of each feature to prediction. Results indicate that XGBoost had good performance in a single run, while the stacked ensemble achieved better performance compared to individual base models, with best performance across all evaluation measures. The system offers actionable recommendations adapted to individual soil nutrient and weather conditions, enabling farmers to make informed crop and fertilizer choices. Through the use of explainable ensemble learning, the proposed solution has implications on sustainable and accurate farming practices.