Large Language Models for Crop Yield Prediction

Abstract Accurate crop yield prediction is vital for agricultural planning and food security. Traditional methods often struggle to integrate diverse datasets, leading to suboptimal predictions. This paper introduces a novel approach leveraging large language models (LLMs), specifically GPT-4, combined with prompt engineering to enhance predictive accuracy. Our method involves crafting specific prompts to guide the LLM in synthesizing data from weather patterns, soil properties, historical yields, and remote sensing. We conducted extensive experiments comparing our method against traditional machine learning models and the Chain of Thought (CoT) approach. Results demonstrate that our method significantly outperforms these baselines in terms of contextual accuracy, explanation quality, and scenario adaptability. This study highlights the potential of LLMs in advancing agricultural analytics and sets the stage for future research in this domain.