Analysis of spatial heterogeneity and influencing factors of carbon emission efficiency at the provincial level in China based on machine learning.

Under increasing pressure from global climate change, improving carbon emission efficiency (CEE) has become an important pathway for promoting green and low-carbon transformation. Using panel data for 30 Chinese provinces from 2006 to 2022, this study develops an integrated analytical framework combining a super-efficiency slacks-based measure (Super-SBM) model with undesirable outputs, spatial autocorrelation analysis, and a random forest model to evaluate CEE and identify its spatiotemporal patterns and key drivers. The results show that: (1) China's CEE increased from 0.275 in 2006 to 0.684 in 2022, although considerable room for improvement remains; (2) CEE exhibits significant spatial clustering, with high-efficiency provinces concentrated along the eastern coast and low-efficiency provinces clustered in the northwest, revealing a clear east-west gradient; (3) GDP per capita, energy intensity, and technology market turnover are the main drivers of CEE; and (4) the driving patterns across China's eight comprehensive economic regions can be grouped into three types: technology-driven, structural transformation-driven, and resource-constrained. This study reveals the spatial heterogeneity of CEE in China within a unified analytical framework. The findings suggest that targeted low-carbon policies should be formulated in accordance with regional differences.