Regional transportation carbon emission prediction and decarbonization pathway analysis based on LASSO-BWO-ELM hybrid heuristic algorithm

Driven by the dual carbon objectives, the transportation sector in Inner Mongolia has embarked on a path toward high-quality growth. Establishing a sustainable, green, and low-carbon transportation network represents a necessary course of action. This study explores the realization of carbon peak and neutrality goals in the transportation industry through forecasting techniques. A hybrid heuristic algorithm combining the Least Absolute Shrinkage and Selection Operator (LASSO), Beluga Whale Optimization (BWO), and Extreme Learning Machine (ELM), namely the LASSO-BWO-ELM model, is used to forecast carbon emissions in Inner Mongolia’s transport industry. Urbanization indicators are included in the model, which combines the Logarithmic Mean Divisia Index (LMDI) and Tapio models to investigate emission drivers and how they decouple from economic growth. This study reduces uncertainty in emission trajectory estimation and policy scenario evaluation by improving forecasting reliability, as higher precision is essential for distinguishing emission pathways and avoiding misjudgments of carbon peak timing and policy effectiveness. The results show superior accuracy compared to other models, achieving MSE = 0.0010, RMSE = 0.0314, MAE = 0.0249, MAPE = 1.59%, and R2 = 99.28% on the test set, thereby providing a reliable quantitative foundation to support future scenario comparison and policy oriented emission reduction research in the transportation sector.