Enhancing Agricultural Predictions using Stacked Ensemble and Transformer-based Models

In the face of growing climate uncertainty and rising food demands, accurate crop yield prediction has become a critical challenge in modern agriculture. Traditional machine learning techniques fail to grasp complex, long-term environmental and soil changes. Fusion of Stacking Ensemble learning and models based on transformers is presented in this study. By combining Decision Trees, XGBoost, and SVR via a neural network meta-learner, the combination improves short-term prediction. Transformer architecture models climatic and soil time-series relationships via self-attention techniques. Compared to current models, the suggested technique showed a 15-20% increase in R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score and decreased prediction errors on real-world datasets. The results suggest that machine learning and deep learning may enhance data-driven sustainable and precision agricultural choices.