Predictive Modeling of Peanut Oil Prices Utilizing a Gaussian Process Regression-Based Machine Learning Framework

Accurate anticipation of fluctuations in commodity valuations is critical for diverse stakeholders, encompassing policymakers, investors, and supply chain entities, to ensure informed decision-making within volatile markets. As a staple edible oil, peanut oil exhibits pronounced price volatility, necessitating robust predictive frameworks to mitigate economic risks. This study leverages a decade-long weekly wholesale price index data set (January 1, 2010–January 10, 2020) to model price dynamics within the Chinese agricultural sector. A Gaussian process regression (GPR) methodology is implemented, integrating Bayesian optimization for hyperparameter tuning and [Formula: see text]-fold cross-validation to systematically evaluate diverse kernel functions and basis configurations. Empirical validation reveals the model’s predictive efficacy, achieving a relative root mean square error (RRMSE) of 0.6823% during the out-of-sample evaluation phase (January 5, 2018–January 10, 2020), underscoring its reliability in capturing nonlinear price trends. The proposed machine learning framework not only serves as an autonomous tool for generating technical price projections but also can complement ensemble forecasting systems by synthesizing insights with econometric or fundamental models. Forecast results here could enhance the granularity of commodity market analyses, offering policymakers and analysts multidimensional perspectives for strategic planning and policy research.