An integrated framework combining deep learning and multicriteria decision-making for SDG-oriented urban low-carbon transition capacity assessment

The Sustainable Development Goals (SDGs) play a critical role in guiding the assessment of Urban Low-Carbon Transition Capacity (ULCTC). However, existing research on ULCTC assessment often overlooks the synergies with the SDGs, as well as spatial-temporal effects and causal relationships among multiple factors within the assessment process. To address these gaps, this study develops an SDG-oriented ULCTC assessment framework and proposes an integrated assessment method that combines deep learning (DL) and multi-criteria decision-making (MCDM) to explicitly incorporate complex attributes into the assessment. Using prefecture-level cities in China as a case study, the results reveal that: (1) the proposed DL model, integrating GCN and LSTM, achieves superior performance in capturing spatial-temporal effects compared to other models; (2) the MCDM approach distinguishes between “cause” and “effect” indicators, and prevents biased assessment outcomes where strong performance in one dimension conceals weaknesses in others, or vice versa; and (3) spatially, ULCTC exhibits strong clustering in regions such as the Beijing-Tianjin-Hebei area, the Yangtze River Delta, and the Pearl River Delta, while temporally, ULCTC shows a steady upward trend, with provincial capitals experiencing the most significant improvements. This study establishes a theoretical bridge between ULCTC and the SDGs, redefines the role of DL in ULCTC assessment, and develops an approach that captures spatio-temporal attributes and causal relationships while balancing synergies and trade-offs, which not only enhances methodological robustness but also generates actionable insights for differentiated policy design, ensuring that low-carbon transition pathways avoid the risks of misjudging short- and long-term priorities or undermining SDG coherence.