Sustainability assessment of cities using multicriteria decision-making combined with deep learning methods

City sustainability is integral to the Sustainable Development Goals for 2030 of the United Nations, and precise assessments are required to inform effective development strategies. However, the assessment of city sustainability is complicated by the conflict and interaction of multiple factors. In this study, we developed a comprehensive model within the triple bottom line (TBL) framework to evaluate the sustainable development performance of 16 cities in Shandong Province, China. Firstly, we integrated spatial-temporal graph convolutional networks combined with long short-term memory (ST-GCN-LSTM) networks to forecast indicator data across economic, social, and environmental dimensions. Secondly, we employed hybrid multicriteria decision-making (MCDM) analysis to evaluate city sustainability performance. Finally, we conducted case studies on the cities of Jinan and Binzhou, proposing targeted improvement strategies based on the influence network relationship map (INRM). The modified Vlse kriterijumska Optimizacija I Kompromisno Resenje (VIKOR) demonstrates that Weihai ranks the highest in terms of sustainability; Jinan should focus on environmental indicators such as green coverage and wastewater management, while Binzhou should prioritize social and environmental factors such as healthcare and social insurance. This study demonstrated that integrating deep learning with MCDM offers a novel and effective approach to city sustainability evaluation.