Application of Intelligence in Agricultural with crop yield prediction

Crop yield prediction is vital for enhancing agricultural productivity and ensuring food security. Machine learning models such as CatBoost, Bagging Regressor, and Ridge Regression have been utilized to predict yields based on historical data, including crop type, seasonal variations, soil conditions, rainfall, and pesticide usage. Among these, CatBoost emerged as the most accurate, achieving an R2 score of 0.954, outperforming Bagging Regressor (R2=0.905) and Ridge Regression (R2=0.400). This superior performance is due to CatBoost’s ability to capture complex, non-linear interactions inherent in agricultural data. Evaluation metrics such as RMSE and MAE further validated CatBoost’s precision, consistent with previous studies reporting up to 99.123Exploratory data analysis revealed key factors influencing crop yield, notably rainfall, temperature, and pesticide application. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) were used to interpret model outputs. SHAP values quantified global feature contributions, while LIME provided local explanations, improving model transparency for stakeholders such as farmers and policymakers. For example, higher rainfall was generally associated with increased yields, whereas extreme temperatures negatively affected output.Future advancements include integrating real-time IoT data to enhance prediction timeliness and exploring deep learning models for greater accuracy. This approach promotes precision agriculture by enabling data-driven decision-making, adaptive resource management, and sustainable farming practices. By combining high predictive power with explainability, AI-driven frameworks like CatBoost contribute significantly to modernizing agriculture and addressing global food challenges.