Advanced Machine Learning Techniques Improve Corn Price Forecasting Accuracy

Strategic Insights -

Recent research conducted by University of Maryland Eastern Shore (UMES) has shed new light on the effectiveness of machine learning techniques in predicting corn prices. The study, led by Raksha Khadka and her team, evaluated the performance of traditional econometric models, as well as advanced machine learning and hybrid approaches, in forecasting corn prices. The results demonstrated that the Neural Network Auto Regressive (NNETAR) model outperformed traditional econometric models, capturing complex temporal structures and adapting to sudden market fluctuations.

Key Takeaways:

  • The study found that traditional econometric forecasting models often fail to capture market volatility, nonlinear trends, and external shocks, making advanced machine learning techniques necessary for accurate forecasting.
  • The NNETAR model emerged as the most effective method for predicting corn prices, outperforming traditional econometric models and adapting to sudden market fluctuations.
  • Hybrid models integrating NNETAR and Seasonal and Trend decomposition using LOESS Model (STLM) demonstrated further improvements in forecasting accuracy, suggesting that combining statistical and machine learning methods enhances predictive reliability.
  • The study used monthly corn price data from 2014 to 2024 to evaluate model performance, providing valuable insights for policymakers to implement market stabilization strategies and adjust trade policies in the U.S. corn sector.
  • The research team consisted of Raksha Khadka, Rumita Limbu Sanwa, Nabin Khadka, and Yeong Nain Chi from UMES.
  • The study's findings underscore the potential of machine learning in agricultural price forecasting, offering a valuable tool for policymakers and stakeholders to inform their decision-making.

Statistics:

  • The study evaluated model performance based on different evaluation metrics, including mean absolute percentage error (MAPE) and root mean squared percentage error (RMSPE).
  • The NNETAR model outperformed traditional econometric models in terms of MAPE and RMSPE values, demonstrating its superiority in capturing complex temporal structures.
  • Hybrid models, combining NNETAR and STLM, showed further improvements in forecasting accuracy, with a 20% reduction in MAPE and 15% reduction in RMSPE compared to the NNETAR model.

Sources:

  • "Forecasting the Global Price of Corn Using Neural Network and Hybrid Approach." IEEE Access, 2025, vol. 13, pp. 167424-167438. (DOI: 10.1109/ACCESS.2025.3611320)
  • University of Maryland Eastern Shore
  • Raksha Khadka, Department of Agriculture, Food and Resource Sciences, University of Maryland Eastern Shore
  • IEEE (Publisher)
  • ScienceDirect (Database provider)