Integrating meteorological and breeding data to predict maize yields using machine learning algorithms

Shaoqiang Wang¹Guangcai Wang²Yucheng Wang¹Yanbing Wang²Huiguang Wang²Weixian Li^3*Jianwei Wei^2*Xuwen Jiang^4*

• ¹Qingdao University of Technology, Qingdao, China
• ²Hebei Academy of Agriculture and Forestry Sciences (HAAFS), Shijiazhuang, China
• ³Langfang Normal University, Langfang, China
• ⁴Qingdao Agricultural University, Qingdao, China

Accurate crop yield prediction is essential for ensuring global food security, especially amid environmental challenges like climate change. While deep learning (DL) techniques have shown promise in yield prediction, they often require large datasets and extensive computational resources. This study develops a machine learning (ML) model to predict maize hybrid yields by integrating meteorological data and breeder-level genetic information, specifically breeding values derived from the best linear unbiased prediction (BLUP) method. We compared four commonly used machine learning algorithms—random forest (RF), XGBoost, support vector regression (SVR), and gaussian process regression (GPR)—and optimized them through hyperparameter tuning. Among these algorithms, the RF algorithm exhibited the best performance, achieving the highest coefficient of determination (R²) of 0.64, along with a low root-mean-square error (RMSE) of 1010.59 kg/ha, mean absolute error (MAE) of 743.89 kg/ha, relative root mean square error (RRMSE) of 10.32%, and mean absolute percentage error (MAPE) of 8.3%. By providing accurate yield predictions for specific cultivars under distinct planting conditions, this predictive model offers valuable support for farmers in selecting well-adapted hybrids, thus promoting more targeted and efficient cultivation practices. Additionally, this study presents a cost-effective and efficient solution for breeders to identify high-yielding maize hybrids optimized for specific environments, ultimately fostering smarter breeding strategies and more precise cultivation recommendations.