Analysis of corn price forecast in China based on Lasso-XGBoost-SHAP

Corn is a pivotal Chinese grain crop , serving as a staple and the largest imported commodity, making accurate price forecasting critical for food security and policy . This study develops an explainable predictive framework for China’s corn prices by integrating supply-demand dynamics, substitute price interactions , and policy factors. Using 2003-2023 monthly data from Chinese and international sources, it employs a two-stage methodology: Lasso (Least Absolute Shrinkage and Selection Operator) regression for feature selection and XGBoost (eXtreme Gradient Boosting) for prediction, the SHAP (Shapley Additive exPlanations) to identify key drivers. The study’s methodological framework and core conclusions are in the below graphical abstract. Results demonstrate that the Lasso-XGBoost model outperforms traditional linear models (LM) and other algorithms, including SVM (Support Vector Machine) and MLP (Multilayer Perceptron), with root mean squared error (RMSE) of 0.094, coefficient of determination (R2) of 0.973, mean absolute error (MAE) of 0.072, representing a 7.84% reduction in RMSE compared to standalone XGBoost. SHAP analysis reveals that medium to late rice prices, China’s corn production and consumption, early indica rice prices, and wheat prices are identified as key determinants . Policymakers should consider the interactions between rice and corn prices and regulate based on domestic supply-demand to stabilize the corn market.