Abstract
Ensuring food security is a critical challenge as the global population and climate instability continue to rise. This study focuses on predicting the yields of corn, soybean, and wheat under these unstable conditions using a machine learning approach. The model utilizes six years of county-level data for five climate variables: temperature, precipitation, solar radiation, CO2, and soil moisture. An Extreme Gradient Boosting (XGBoost) model was trained for each crop, and the SHAP (SHapley Additive exPlanations) framework was employed to ensure model interpretability.
Separate models were built for each crop, with the corn and soybean models demonstrating high predictive accuracy, achieving R² values of 0.76 and 0.81, respectively. In contrast, the wheat model yielded a lower R² value of approximately 0.5. A SHAP analysis of the high-performing corn and soybean models revealed that temperature and precipitation during the summer growing season were the most influential predictors, a finding that aligns with established agronomic principles.
The ability to predict the yield of major food crops is a critical step toward ensuring food security, especially under unstable climate conditions. The findings from this study can empower governments to formulate data-driven agricultural policies for the future, while also providing farmers with new, data-backed insights to consider when creating their planting plans.
Advisor
Pasteur, Drew
Department
Statistical and Data Sciences
Recommended Citation
Lee, Yongchan, "Predicting Crop Yield Using Climate Variables with a Machine Learning Approach" (2025). Senior Independent Study Theses. Paper 11753.
https://openworks.wooster.edu/independentstudy/11753
Disciplines
Agronomy and Crop Sciences | Data Science | Environmental Sciences
Keywords
Crop Yield Prediction, Food Security, Machine Learning, Remote Sensing, SHAP
Publication Date
2025
Degree Granted
Bachelor of Arts
Document Type
Senior Independent Study Thesis
© Copyright 2025 Yongchan Lee