Forecasting the total carbon allowance cap under emission-reduction targets using a hybrid path analysis and supervised machine learning framework

Carbon allowances constitute a foundational component of national carbon emission control frameworks, as they govern the equitable distribution of subsequent allocations and directly shape the overall effectiveness of greenhouse gas mitigation strategies. However, the temporal evolution of carbon allowances is inherently complex, high-dimensional, and nonlinear, thereby posing substantial challenges to the rigorous prediction of the aggregate allowance cap. Although artificial intelligence technologies have achieved substantial advances in environmental forecasting in recent years, existing predictive approaches often prioritize predictive accuracy while neglecting systematic variable selection and structured modeling procedures, thereby constraining their utility for policy-oriented decision support. To address this limitation, we propose a hybrid modeling framework that integrates path analysis with supervised machine learning to forecast China’s future carbon allowance cap. Path analysis was first applied to disentangle both direct and indirect relationships among the variables, thereby enabling the identification of structurally significant predictors with substantial explanatory power. Based on the selected indicators, a standardized dataset was constructed to train and systematically compare multiple supervised machine learning algorithms. Empirical results demonstrate that, under uniformly regularized data conditions, Gaussian Process Regression (GPR) consistently outperforms alternative supervised learning algorithms in terms of predictive accuracy and robustness. The integrated forecasting framework developed herein provides a robust analytical foundation for identifying the determinants of carbon allowance trajectories and illustrates how machine learning can be effectively integrated with environmental datasets to inform carbon governance, strengthen climate mitigation pathways, and advance data-driven environmental decision-making under realistic emission reduction targets.