Work

Application of Machine Learning Methods to Predict Corn Harvest Yields Based on Climate Data

Year: 2025
Type: article
Abstract: Purpose: The study aims to evaluate the effectiveness of machine learning (ML) methods in predicting corn yields under climate variability, addressing the limitations of traditional statistical models in capturing nonlinear and dynamic crop environment interactions. Subjects and Methods: Machine learning algorithms including Random Forest (RF), Gradient Boosting Machines (GBM), Gaussian Process Regression (GPR), and Support Vector Regression (SVR) were applied to datasets comprising climatic, soil, and vegetation index (VI) variables. Model performance was assessed using standard evaluation metrics such as the coefficient of determination (R²), root mean square error (RMSE), and normalized RMSE (nRMSE). Comparative analyses were conducted across different crop growth stages (V1–R6). Results: Ensemble and hybrid models outperformed single algorithms, with GBM achieving the highest overall accuracy (R² ≈ 0.85; RMSE ≈ 0.45 t/ha). RF consistently served as a robust baseline across datasets. Multimodal integration of VIs, soil, and climatic variables significantly improved accuracy, particularly during early growth stages where VI-only models underperformed. At maturity, GPR and RF achieved strong performance (RMSE ≈ 1.80 Mg/ha; nRMSE ≈ 13.5%). SVR demonstrated resilience under conditions of reduced data availability, making it effective for in-season forecasts. Conclusions: Machine learning provides a powerful and adaptive framework for corn yield prediction. By integrating diverse datasets and leveraging ensemble and hybrid models, forecasting accuracy can be improved for both early-season decision-making and end-of-season yield estimation. These results highlight the potential of ML to enhance agricultural resilience and inform climate adaptation strategies.
Author: Agung, Andi
Language: English
Subfield: Plant Science
Open Access status: diamond
OpenAlex ID: W7220022533
FWCI:
Cites: 19
Cited by:
Related to: 10