Sustainable regional corn yield prediction for the United States through interpretable machine learning approach.

Considering crop yield prediction as critical to optimizing agricultural practices and food security, this question is critical to U.S. agricultural planning and regional food security; relevant research on corn, one of the essential crops, must focus on the accurate methods for predicting this crop. It has been discussed that yield prediction models generally rely on simplistic approaches, which fail to capture complex, non-linear relationships in agricultural data. This work fills the knowledge gap by making use of advanced machine-learning techniques to improve the accuracy of corn yield prediction. This study focuses on county-level regional forecasting(U.S) to support agricultural policy and supply chain planning rather than field-specific management decisions. The methodology is in line with the Special Section on Sustainable Computing for Next-Generation Low-Carbon Agricultural Consumer Electronics by designing a data-efficient algorithm that focuses on the Random Forest Classifier, Gradient Boosting Classifier, and Ensemble Voting Classifier. The development of this model entailed the pre-processing of historical data concerning corn yield, defining pertinent attributes, and assessing the confusion matrix, ROC curve, and SHAP values for explainability. This work proposes an ensemble model which has achieved remarkable accuracy and robustness, excelling in performance relative to the existing approaches. The model has also made solid predictions, with a precision, recall, and F1-score of 0.92 and a training accuracy of 0.97. The SHAP further enhances transparency into the features that drive predictions, hence making the model more interpretable. This is of great importance to agricultural planning; this would most probably offer a sound instrument to predict corn yield and optimize resources in agricultural consumer practices. This paper strongly advocates energy-efficient algorithm design, intelligent applications, sustainable computing, efficiency, environmental impact, and resource recycling to drive toward sustainable and efficient corn yield prediction.