Interpretable Crop Selection for Optimized Farming Decisions

: The success of agriculture depends on e ff ective crop selection, which influences yield, profitability, and risk management for farmers. Although machine learning tools are increasingly used for crop recommendations, many current models operate as opaque ”black boxes”, causing farmers to hesitate due to a lack of transparency. This study introduces an interpretable crop selection model that leverages the AdaBoost classifier, using soil and climate data to predict crop suitability. To ensure transparency and foster trust, we incorporate SHapley Additive Explanations (SHAP) to break down the model’s decision-making process. SHAP plots visually illustrate how each input such as nitrogen, phosphorus, potassium, pH, temperature, humidity, and rainfall contributes to crop predictions. These visual aids o ff er farmers practical, actionable insights, helping them understand the rationale behind the system’s recommendations. Our model was evaluated on a dataset of 22 crops, achieving outstanding accuracy (99.77%) with a rapid prediction time of 0.5 seconds per query. Transparency is provided not only through SHAP visualizations but also through clear, user-friendly interfaces that display feature contributions in an accessible manner. This combination of high predictive performance and easy-to-interpret explanations empowers farmers to make informed, confident decisions, leading to improved crop yields and greater profitability.