A crop recommendation using improved expert-guided constraint aware differential evolution and CatBoost surrogate...

Planning for sustainable crops, among other things, depends on smart decision-support systems capable of managing multi-criteria interactions and different environmental conditions. This study presents a multi-criterion, scenario-aware crop recommendation system developed by coupling Improved ECA-DE (Expert-guided Constraint-Aware Differential Evolution) optimization with CatBoost (Categorical Boosting) surrogate learning. The system assesses the suitability of crops using a range of agro-environmental factors, such as soil type, soil pH, temperature, humidity, water requirement, crop duration, and water source under three scenarios: balanced, drought, and short-season conditions. Improved ECA-DE is used for the determination of criterion weights, and the CatBoost model approximates expert-based suitability scores very effectively. The prediction performance of the proposed method was very high, with coefficient of determination (R ²) = 0.9975, Mean Absolute Error (MAE) = 0.0043, and Root Mean Square Error (RMSE) = 0.0056, showing almost perfect matches between the predictions and expert judgments. In addition, the Improved ECA-DE can drastically lower the computation time and memory usage while providing Standard DE matching in terms of optimization quality. In addition, the scenario-based assessment revealed that the method can accurately represent the trade-offs in suitability, water requirement, and crop duration, thus facilitating adaptive recommendations under different conditions. The proposed framework is a precise, interpretable, and resource-efficient tool for smart and sustainable agricultural decision-making.