A Carbon Emission Prediction Method for Sustainable Transportation: An ENR–BAS–GRNN Hybrid Computational Intelligence Framework

Accurate prediction of transportation carbon emissions is essential for identifying emerging emission pressures and supporting the transition toward sustainable transportation systems. This study develops a hybrid prediction framework that integrates Elastic Net Regression (ENR), Beetle Antennae Search (BAS), and Generalized Regression Neural Network (GRNN) to estimate transportation-sector monthly carbon emissions from 2008 to 2023 in Anhui Province, China. ENR is employed to identify key influencing factors, while BAS optimizes GRNN parameters to enhance model convergence and generalization. In an ex-post out-of-sample evaluation using observed explanatory variables from 2023, comparative experiments show that the ENR–BAS–GRNN model achieves the highest accuracy (R2 = 0.99, ARE < 1%), outperforming all benchmark models. The results indicate that integrating feature selection with intelligent optimization can effectively capture nonlinear emission dynamics. Beyond improving estimation accuracy, the proposed framework provides a quantitative tool for monitoring transportation-related carbon emissions and diagnosing the factors associated with emission changes. When coupled with scenario-based projections of explanatory variables, the framework can support ex-ante forecasting. The study therefore contributes a computationally lightweight and data-driven approach for integrating carbon emission measurement, monitoring, and policy support into regional sustainable transportation governance.